Target allocation method based on improved particle swarm optimization
By optimizing the iterative parameter update method in the particle swarm algorithm, the particle swarm algorithm is easily trapped in the local optimal problem in the target allocation problem, achieving more reasonable target allocation and more efficient drone cluster strike effect.
Patent Information
- Application Number
- CN202411886535.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-20
- Publication Date
- 2025-04-29
AI Technical Summary
Particle swarm algorithms are prone to falling into local optimization in the target allocation problem, resulting in unreasonable allocation in the scenario of attacking drone cluster missions.
By optimizing the update method of iterative parameters, a target allocation method based on improved particle swarm algorithm is designed, including adjusting the inertial weight, acceleration constant and maximum speed update method to avoid local optimal traps.
It effectively avoids the problem of target allocation algorithm falling into local optimality too quickly, provides a more reasonable target allocation plan, and improves the overall strike efficiency of ground-air firepower on UAV cluster targets.
Smart Images
Figure CN119903869A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of operations research and relates to a target allocation method based on an improved particle swarm algorithm. Background Art
[0002] Countering drone swarms is currently a research hotspot for countries around the world. In addition to adding various active / passive defense methods for ground targets, the most effective countermeasure against drone swarms is to destroy them at a long distance with surface-to-air firepower.
[0003] The core idea of the particle swarm optimization algorithm is to use the information sharing mechanism between individuals to enable the entire group to evolve from disordered movement to ordered movement in the problem-solving space, thereby obtaining a feasible solution to the optimization problem.
[0004] However, in practical applications, since particles constantly track the current optimal solution during the iteration process, the particle population has significant convergence in the later stages of the iteration. The lack of diversity causes the particle swarm algorithm to easily fall into the local optimum. In the scenario of combating drone swarms, the particle swarm algorithm is prone to fall into the local optimum when solving the target allocation problem, which needs to be improved. Summary of the invention
[0005] In view of the above-mentioned defects of the prior art, the present invention is based on the particle swarm algorithm, and designs a target allocation method based on the improved particle swarm algorithm by optimizing the updating method of iterative parameters, so as to provide a more reasonable target allocation scheme for the mission scenario of combating drone clusters.
[0006] The technical solution adopted by the present invention to solve the technical problem is: a target allocation method based on an improved particle swarm algorithm, the steps are as follows:
[0007] S1, set the population size and iteration times, and initialize the position and velocity of each particle;
[0008] S2, calculate the fitness of each particle, record the pbest of each particle and the current gbest of the population;
[0009] S3, update the position and velocity of the particle;
[0010] S4, update inertia weight;
[0011] S5, update acceleration constant;
[0012] S6, compare the fitness value with the historical optimal position of the particle, store the current optimal position of each particle in pbest, and update gbest;
[0013] S7, determine whether the iteration is terminated; if not, repeat step S3; if yes, output the optimal solution.
[0014] Furthermore, in the step S1, at the beginning, the particle swarm algorithm first initializes a group of randomly distributed particles, and in subsequent iterations, continuously tracks the optimal particles to explore the solution space until the optimal solution is found.
[0015] Furthermore, in each iteration of step S3, each particle is assigned the optimal position of the individual particle and the optimal position of the group particle by the formula Update its own position and velocity in the next iteration, where v(t), x(t), p best (t), g best (t) are the particle velocity, particle position, individual optimal position, and group optimal position at the tth iteration, ω represents the inertia weight, c 1 、c 2 represents the acceleration constant associated with different impact factors, r 1 、r 2 Represents a random number in the range [0,1].
[0016] Furthermore, the particle speed is updated in step S3 as follows: in each round of iterative update, the moving speed of each particle in the population is within a fixed interval [v min ,v max ], usually satisfy v max =-v min , when v max A larger value is beneficial for particles to jump out of the local optimum in the initial stage and enhance the global search ability, but it will also cause the search range of particles to be too large in the later iteration, affecting the local convergence of the algorithm. Therefore, the ideal maximum speed v max It should gradually decrease as the number of iterations increases, as shown in the following formula: where v max (t) represents the maximum speed in the current iteration round, t represents the current iteration number, n represents the maximum number of iterations of the algorithm, and v max 、v min are the initial velocity value boundaries respectively.
[0017] Further, the step S4 is specifically: adopting a nonlinear decreasing value change mode, as shown in the following formula: where ω start ,ω end are the inertia weights of the start and end rounds of iteration, t represents the current number of iterations, and n represents the maximum number of iterations of the algorithm. According to the monotonicity of the tangent function, it can be ensured that the value of ω(t) in the iteration process is within [ω start ,ω end ]It decreases nonlinearly within the interval.
[0018] Further, the step S5 is specifically as follows: the acceleration constant c1 、c 2 This allows particles to move closer to the optimal solution in each iteration. 1 Need to gradually decrease during the iteration process, and c 2 Then it gradually increases during the iteration process, and the acceleration constant c 1 、c 2 Using a differentiated gradient strategy can achieve better performance, as shown in the following formula: where c 1start 、c 1end They are the acceleration constants c 1 At the beginning and end of the iteration, the value of c 1start >c 1end , c 2start 、c 2end They are the acceleration constants c 2 At the beginning and end of the iteration, the value of c 2start <c 2end , t represents the current iteration number, and n represents the maximum iteration number of the algorithm.
[0019] The beneficial effects of the present invention are as follows: compared with the original particle swarm optimization algorithm, the present invention can effectively avoid the problem of the target allocation algorithm falling into the local optimum too quickly, and provide a more reasonable target allocation scheme for the target allocation problem. The present invention designs a target allocation method based on the particle swarm algorithm, which can improve the overall strike effectiveness of ground-to-air firepower against drone cluster targets. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 Flowchart of the method for allocating objects of the present invention. DETAILED DESCRIPTION
[0021] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.
[0022] Reference Figure 1 As shown, the present invention discloses a target allocation method based on an improved particle swarm algorithm, and the steps are as follows.
[0023] S1, set the population size and number of iterations, and initialize the position and velocity of each particle.
[0024] At the beginning, the particle swarm algorithm first initializes a group of randomly distributed particles, and in subsequent iterations, it continuously tracks the optimal particles to explore the solution space until the optimal solution is found.
[0025] In each iteration, each particle updates its position and velocity in the next iteration according to the current individual best position and group best position, as shown in the following formula:
[0026]
[0027] where v(t), x(t), p best (t), g best (t) are the particle velocity, particle position, individual optimal position, and group optimal position at the tth iteration, ω represents the inertia weight, c 1 、c 2 represents the acceleration constant associated with different impact factors, r 1 、r 2 Represents a random number with a value range of [0,1]. According to the above formula, the update result of the particle velocity position is mainly related to the inertia weight ω, particle velocity v(t), acceleration constant c 1 、c 2 These are related to other parameters, so they are improved respectively.
[0028] S2, calculate the fitness of each particle, record the pbest of each particle and the current gbest of the population.
[0029] S3, update the position and velocity of the particle.
[0030] Improved particle speed update range: In each round of iterative update, the movement speed of each particle in the population is in a fixed range [v min ,v max ], usually satisfy v max =-v min In the particle velocity range, the main parameter affecting the convergence of the algorithm is the maximum velocity v max :When v max When the value is large, it is beneficial for particles to jump out of the local optimum in the initial stage and enhance the global search ability, but it will also cause the search range of particles to be too large in the later iteration, affecting the local convergence of the algorithm. Therefore, the ideal maximum speed v max It should gradually decrease as the number of iterations increases, as shown in the following formula: where v max (t) represents the maximum speed in the current iteration round, t represents the current iteration number, n represents the maximum number of iterations of the algorithm, and v max 、v min are the initial speed value boundaries respectively. According to the analysis, the maximum speed v max The initial value of (t) is v max , and gradually decreases with the increase of the number of iterations, meeting the performance requirements.
[0031] S4, update the inertia weight.
[0032] Improved inertia weight update method: The inertia weight ω determines the inheritance of the current velocity of the particle in the next iteration process, so it has a significant impact on the convergence performance of the particle swarm algorithm: a larger ω value can improve the global search ability of the particles and maintain the diversity of the particle population; a smaller ω value is conducive to local search of particles and accelerates the convergence of the algorithm.
[0033] In summary, the reasonable ω value should decrease with the increase of iteration rounds. In the choice of decreasing method, since the linear decreasing inertia weight ω is not easy to get rid of the local optimum, it is necessary to adopt a nonlinear decreasing value change method, as shown in the following formula: where ω start ,ω end are the inertia weights of the start and end rounds of iteration, t represents the current number of iterations, and n represents the maximum number of iterations of the algorithm. According to the monotonicity of the tangent function, it can be ensured that the value of ω(t) in the iteration process is within [ω start ,ω end ]It decreases nonlinearly within the interval.
[0034] S5, update the acceleration constant.
[0035] Improved acceleration constant update method: acceleration constant c 1 、c 2 They enable particles to move closer to the optimal solution in each round of iteration. They determine the influence of the particle's own best experience and the group's best experience, that is, the self-learning ability and the group learning ability: the particle's self-learning helps to enhance the global search ability, while the group learning helps to enhance the local convergence. Therefore, c 1 Need to gradually decrease during the iteration process, and c 2 Then it gradually increases during the iteration process, and the acceleration constant c 1 、c 2 Using a differentiated gradient strategy can achieve better performance, as shown in the following formula: where c 1start 、c 1end They are the acceleration constants c 1 The value of the iteration start and end rounds, and c 1start >c 1end ; and c 2start 、c 2end They are the acceleration constants c 2 The value of the iteration start and end rounds, and c 2start <c 2end , t represents the current iteration number, n represents the maximum iteration number of the algorithm
[0036] S6, compare the fitness value with the historical optimal position of the particle, store the current optimal position of each particle in pbest, and update gbest.
[0037] S7, determine whether the iteration is terminated; if not, repeat step S3; if yes, output the optimal solution.
[0038] The present invention designs a target allocation method based on a particle swarm algorithm, which can improve the overall strike effectiveness of ground-to-air firepower against UAV cluster targets.
[0039] The above embodiments are only illustrative of the principles and effects of the present invention, as well as some embodiments of its application. For those skilled in the art, several modifications and improvements may be made without departing from the creative concept of the present invention, and all of these belong to the protection scope of the present invention.
Claims
1. A target allocation method based on an improved particle swarm algorithm, characterized in that: The steps are as follows S1, set the population size and iteration times, and initialize the position and velocity of each particle; S2, calculate the fitness of each particle, record the pbest of each particle and the current gbest of the population; S3, update the position and velocity of the particle; S4, update inertia weight; S5, update acceleration constant; S6, compare the fitness value with the historical optimal position of the particle, store the current optimal position of each particle in pbest, and update gbest; S7, determine whether the iteration is terminated; if not, repeat step S3; if yes, output the optimal solution.
2. The target allocation method based on improved particle swarm algorithm according to claim 1 is characterized in that: In the step S1, a group of randomly distributed particles are first initialized, and in subsequent iterations, the optimal particles are continuously tracked to explore the solution space until the optimal solution is found.
3. The target allocation method based on improved particle swarm algorithm according to claim 2 is characterized in that: In each iteration of step S3, each particle is assigned the optimal position of the individual particle and the optimal position of the group particle by the formula Update its own position and velocity in the next iteration, where v(t), x(t), p best (t), g best (t) are the particle velocity, particle position, individual optimal position and group optimal position at the tth iteration respectively, ω represents the inertia weight, c1 and c2 represent the acceleration constants associated with different influencing factors, and r1 and r2 represent random numbers in the range of [0,1].
4. The target allocation method based on improved particle swarm algorithm according to claim 3 is characterized in that: The particle speed is updated in step S3 as follows: the moving speed of each particle in the population in each round of iterative update is within a fixed interval [v min ,v max ] and satisfies v max =-v min ; Maximum speed v max As the number of iterations increases, it gradually decreases: where v max (t) represents the maximum speed in the current iteration round, t represents the current iteration number, n represents the maximum number of iterations of the algorithm, and v max 、v min are the initial velocity value boundaries respectively.
5. The target allocation method based on improved particle swarm algorithm according to claim 4 is characterized in that: The step S4 is specifically as follows: based on the formula Adopt nonlinear decreasing value change mode, where ω start ,ω end are the inertia weights of the start and end rounds of iteration, t represents the current number of iterations, and n represents the maximum number of iterations of the algorithm. It is ensured that the value of ω(t) during the iteration process is within [ω start ,ω end ]It decreases nonlinearly within the interval.
6. A target allocation method based on an improved particle swarm algorithm according to any one of claims 1 to 5, characterized in that: The step S5 is specifically as follows: based on the formula A differentiated gradient strategy is adopted for the acceleration constants c1 and c2, where c 1start 、c 1end are the values of the acceleration constant c1 at the beginning and end of the iteration, and c 1start >c 1end , c 2start 、c 2end are the values of the acceleration constant c2 at the beginning and end of the iteration, and c 2start <c 2end , t represents the current iteration number, and n represents the maximum iteration number of the algorithm.