Improved genetic algorithm-based task allocation method for swarm unmanned aerial vehicle system

Abstract

The invention provides a swarm unmanned aerial vehicle system-oriented task allocation method based on an improved genetic algorithm, and the method comprises the steps: building a task target library, quantifying the damage capability equivalent value of a single unmanned aerial vehicle as 1, and determining the anti-damage capability equivalent value of each target in the task target library; constructing a task allocation efficiency evaluation model; constructing a task allocation matrix by taking the target number and the unmanned aerial vehicle number as rows and columns; in the flying process of the swarm unmanned aerial vehicles, the unmanned aerial vehicles discovering the targets perform bid invitation; the bidding unmanned aerial vehicle performs chromosome gene coding according to the data chain numbering sequence; randomly generating an initial population for genetic algorithm processing, calculating a fitness function for each generation of individuals, and selecting individuals with fitness meeting a set threshold value for next generation of genetic processing; and decoding the optimal individual to obtain a task allocation matrix until the maximum genetic algebra is met. The method is suitable for rapidly completing dynamic task allocation under various different scenes and various uncertain conditions.

Description

Technical Field

[0001] The invention belongs to the technical field of multi-agent task allocation, and in particular relates to a dynamic task allocation method based on an improved genetic algorithm suitable for distributed bee colonies. Background Art

[0002] Swarm drones have the advantages of low cost and distributed distribution, but they also have certain limitations: the drones themselves have a short range and can only perform one strike mission.

[0003] The traditional task allocation method is based on the ability of drones to perform tasks. One drone can perform multiple strike missions, so it is a drone-centric task allocation method. Swarm drones are fundamentally different from large drones in terms of range and function: they have a short range and can only perform one strike mission. At the same time, the mission target requires multiple swarm drones to strike in coordination, so there is an urgent need for a task allocation method that is suitable for swarm drone systems and centered on mission targets.

[0004] In addition, in the actual battlefield, there may be situations where the mission targets move, there may also be situations where the swarm UAV system discovers new mission targets, and there may be situations where the swarm UAVs are destroyed by the enemy when approaching the mission area. The dynamic uncertainties in the combat area bring huge challenges to the swarm UAV system mission. In the ever-changing battlefield environment, swarm UAVs need to complete task allocation quickly. The swarm UAV system needs to perform dynamic task allocation when the task bidding UAV number is uncertain, the UAV bidding location is uncertain, the number of task targets in the task cluster is uncertain, and the task target equivalent value is uncertain. Summary of the invention

[0005] The present invention aims to solve one of the above-mentioned technical problems and provides a task allocation method for a swarm UAV system based on an improved genetic algorithm. By quantifying the task objectives, a task allocation matrix is ​​constructed, a genetic coding method suitable for swarm UAVs is designed, and task allocation is performed based on an improved genetic algorithm. This method can quickly complete dynamic task allocation in various scenarios and under various uncertain conditions.

[0006] The technical solution adopted by the present invention to solve the above technical problems is as follows:

[0007] The present invention provides a task allocation method for a swarm UAV system based on an improved genetic algorithm, comprising the following steps:

[0008] Establish a mission objective library of typical targets in the combat area;

[0009] Quantify the destructive power of the swarm drones used, and set the destructive power equivalent value of a single drone to 1;

[0010] According to the damage capability equivalent value of a single UAV, determine the anti-damage capability equivalent value of each target in the mission target library;

[0011] The task allocation efficiency evaluation model is constructed based on the difference between the total target value benefit and the total cost;

[0012] Construct a task allocation matrix with the number of targets as rows and the number of drones as columns;

[0013] The drone that finds the target during the flight of the swarm drone acts as the auctioneer and conducts bidding;

[0014] The bidding drones will use integer arrangement coding to encode chromosome genes according to the data link number sequence;

[0015] Randomly generate an initial population as the starting solution, perform genetic algorithm processing, calculate the fitness function for each generation of individuals, select individuals whose fitness meets the set threshold, and perform genetic processing for the next generation;

[0016] Until the maximum genetic generation is met, the best individual is decoded to obtain the optimal task allocation matrix.

[0017] Furthermore, the anti-damage capability equivalent value of the target is determined through combat examples, actual combat exercises or simulation verification methods.

[0018] Furthermore, the task allocation efficiency evaluation model is

[0019] U=Reward-PathCost-RiskCost

[0020]

[0021]

[0022]

[0023] Where i represents the drone number, j represents the target number, n represents the total number of drones participating in the bidding, and m represents the total number of targets in the area; g j represents the jth target, represents the i-th drone that attacks target j, represents the position of UAV i that is executing the attack on target j, pg j represents the position of target j; w j represents the equivalent value of the jth target; q j represents the profit coefficient of the jth target; β j represents the risk factor of the jth target; Reward is the value of the target, Reward(g j) represents the value of target j, PathCost is the voyage cost, PathCost(pu i ,pg j ) represents the flight cost from UAV i to target j; RiskCost is the risk cost, RiskCost(pu i ,pg j ) represents the risk of UAV i being destroyed on its way to target j; U represents the task allocation efficiency.

[0024] Furthermore, the fitness function is:

[0025]

[0026] Among them, U represents the task allocation efficiency, s represents the sth individual in the population, and f(s) represents the fitness of the sth individual.

[0027] Furthermore, the task allocation efficiency evaluation model is

[0028]

[0029] Among them, j represents the target number, and m represents the total number of targets in the area; represents the position of UAV i that is executing the attack on target j, pg j represents the position of target j; w j represents the equivalent value of the jth target, PathCost(pu i ,pg j ) represents the flight cost from UAV i to target j; U represents the task allocation efficiency.

[0030] Furthermore, the fitness function is:

[0031]

[0032] Among them, s represents the sth individual in the population, f(s) represents the fitness of the sth individual, and t represents the sum of all target equivalent values. is the distance from the i-th UAV to the j-th mission target.

[0033] Furthermore, the swarm of drones uses a contract network protocol to exchange information, adopts a distributed control method, each drone performs cost calculation in a distributed manner, and the auctioneer completes and sends the allocation results in a centralized manner.

[0034] Furthermore, the chromosome includes a dominant gene and a recessive gene, the dominant gene represents a drone that performs a strike mission, and the recessive gene represents a drone that does not perform a strike mission.

[0035] Furthermore, the genetic treatment comprises

[0036] Select individuals with high fitness in the population to enter the next generation;

[0037] The parent individuals are grouped in pairs for crossover operation;

[0038] Parent individuals undergo random mutation operations according to a certain ratio;

[0039] The parent individual performs a reversal operation.

[0040] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the task allocation method.

[0041] The beneficial effects of the present invention compared with the prior art are as follows:

[0042] (1) Considering the fact that swarm drones can only perform one strike mission and the mission target requires the coordinated strike of multiple swarm drones, the traditional drone-centered task allocation method is changed to a task target-centered allocation method, namely, a target equivalent-based task allocation method.

[0043] (2) Based on the fact that there are bidders in the task allocation process of the swarm UAV system, combined with the characteristics of data link numbering in the swarm UAV system, dominant genes and recessive genes are designed in the genetic algorithm, and a gene encoding method suitable for distributed swarms is proposed;

[0044] (3) In order to solve the problem that uncertain factors in the combat area will affect task allocation, a dynamic task allocation method based on an improved genetic algorithm is proposed. The task allocation method based on the improved genetic algorithm can quickly find a more optimal allocation strategy. The dynamic task allocation method based on the improved genetic algorithm can complete dynamic task allocation when the bidding drone number is uncertain, the bidding drone position is uncertain, the number of task targets is uncertain, and the task target equivalent value is uncertain. It has high dynamic adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The included drawings are used to provide a further understanding of the embodiments of the present invention, which constitute a part of the specification, are used to illustrate the embodiments of the present invention, and together with the text description, explain the principles of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 A schematic diagram of the strike distribution principle based on target equivalence provided by a specific embodiment of the present invention;

[0047] Figure 2 A flow chart of a task allocation method based on an improved genetic algorithm provided in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0048] Specific embodiments of the present invention are described in detail below. In the following description, for the purpose of explanation and not limitation, specific details are set forth to help fully understand the present invention. However, it will be apparent to those skilled in the art that the present invention may also be practiced in other embodiments that depart from these specific details.

[0049] It should be noted that in order to avoid obscuring the present invention due to unnecessary details, only the device structure and / or processing steps closely related to the scheme of the present invention are shown in the drawings, while other details that are not closely related to the present invention are omitted.

[0050] As one aspect of the present invention, a task allocation method for a swarm drone system based on an improved genetic algorithm is provided, comprising the following steps:

[0051] S1: Establish a mission target library of typical targets in the combat area

[0052] Nowadays, countries have a wide variety of weapons and equipment, and the mobility, strike benefits, threat performance, and damage resistance of different equipment are different; secondly, there is no target recognition algorithm that can identify global equipment; thirdly, the types of weapons owned by specific areas and specific combat targets are relatively fixed (such as HIMARS and F22). Therefore, it is necessary to establish a typical mission target library for specific combat areas and specific combat targets.

[0053] S2: Quantify the destructive capability of this type of swarm drone, and define the destructive capability equivalent value of a drone as 1

[0054] Different swarm drones have different carrying capacities and therefore different destructive capabilities. The damage equivalent value is defined for the destructive capability of this type of swarm drone. The equivalent value of the swarm drone is defined, and based on this, each mission target in the mission target library is modeled for anti-damage equivalent, so that the algorithm is applicable to different application scenarios and has universality.

[0055] S3: Based on the damage capability of this type of swarm drone, the damage resistance of typical mission targets in the combat area is quantitatively modeled

[0056] Taking an armored vehicle as an example, the damage resistance equivalent value of 3 means that three swarm drones are needed to completely destroy the armored vehicle, and the damage resistance equivalent value is rounded up.

[0057] The anti-damage equivalent value of the mission target is mainly obtained through three methods: combat examples, actual combat exercises and simulation verification.

[0058] (1) The combat examples are derived from actual combat results, that is, the ammunition equivalent of the swarm UAVs to a certain type of ground-based air defense firepower system is determined by the actual combat effect of the swarm UAVs on the ground-based air defense weapons in actual combat, and the results are highly reliable.

[0059] (2) The actual combat exercise is derived from the results of air-to-ground attack and defense exercises. That is, the equivalent value of the swarm of drones to the mission target is determined through the exercise. It is highly targeted and can practice various attack tactics to obtain more comprehensive and reliable analysis results.

[0060] (3) Simulation verification comes from air-to-ground attack and defense simulation, that is, the equivalent value of the swarm of drones to the mission target is determined through simulation. The accuracy of the analysis results is affected by the combat environment, combat assumptions, combat tactics and the authenticity of the simulation of model building. An unlimited number of simulation confrontations can be carried out, which is only limited by computer performance and running time. The cost is small and the results are relatively reliable.

[0061] The present invention assumes that the ammunition equivalents required to strike different targets can be obtained from a database, and the equivalent values ​​of different mission targets are known.

[0062] S4: Constructing a UAV bidding performance evaluation model (performed by bidders)

[0063] The bidding performance evaluation function is the revenue of the UAV completing the task minus the corresponding cost. The revenue is mainly the value of the target Reward (g j ), the cost mainly includes the voyage cost PathCost (pu i ,pg j ) and risk cost RiskCost(pu i ,pg j ), the risk cost represents the risk of the drone being destroyed during the mission, and the bidding effectiveness function is used to calculate the effectiveness of a single drone against different targets:

[0064] U(pu i ,pg j )=α×Reward(g j )-PathCost(pu i ,pg j )-β×RiskCost(pu i ,pg j )

[0065] in, q represents the profit coefficient of the target, Indicates w j The value of the profit after destroying the target is evenly shared among the two drones, and β represents the risk coefficient.

[0066] It should be noted that, in some embodiments, the designed scenario is to destroy all targets, the total target value benefit can be equivalent to a constant value, and the risk cost is ignored to simplify the calculation model, so the bidding performance function can be equivalent to the solution path cost.

[0067] S5: Construct task allocation performance evaluation model (performed by the tenderer)

[0068] The task allocation efficiency evaluation model aims to maximize the global benefit. The task allocation efficiency evaluation model is composed of the total target value reward minus the total cost. The total cost includes the total voyage cost PathCost and the total risk cost RiskCost:

[0069]

[0070]

[0071]

[0072] U=Reward-PathCost-RiskCost

[0073] Where i represents the drone number, j represents the target number, n represents the total number of drones participating in the bidding, and m represents the total number of targets in the area; g j represents the jth target, u i represents the i-th drone, represents the i-th UAV that executes the attack on target j, pu i Indicates the position of drone i, pg j represents the position of target j; w represents the target equivalent value, w j represents the equivalent value of the jth target; q represents the profit coefficient of the target, q j represents the profit coefficient of the jth target; β represents the risk coefficient of the path, β j represents the risk factor of the jth target; Reward is the value of the target, Reward(g j ) represents the value of target j, PathCost is the voyage cost, PathCost(pu i ,pg j ) represents the flight cost from UAV i to target j; RiskCost is the risk cost, RiskCost(pu i ,pg j ) represents the risk of UAV i being destroyed on its way to target j; U represents the task allocation efficiency.

[0074] It should be noted that in some embodiments, the designed scenario is to destroy all targets, the total target value benefit can be equivalent to a constant value, and the risk cost is ignored to simplify the calculation model. Therefore, the total task allocation efficiency function can be equivalent to solving the minimum path cost, that is,

[0075]

[0076] S6: Construct task allocation matrix A

[0077] The rows of the matrix represent the mission targets discovered by the swarm drone system, and the columns of the matrix represent the drones participating in the current bidding mission. The number of rows in the task allocation matrix is ​​the number of targets discovered, and the number of columns in the task allocation matrix is ​​the number of drones participating in the bidding. A(j,i)=1 means that the i-th drone will attack the j-th target, and the sum of the rows is equal to the equivalent value w of the target. j , the sum of the columns is less than or equal to 1 (1 means that the drone is assigned a strike mission, and 0 means that the drone is a bidder for this mission).

[0078] S7: Construct the range cost matrix D between the drone and the target

[0079] The dimension of the range cost matrix D is the same as that of the task allocation matrix A. The rows of the matrix represent the task targets discovered by the swarm UAV system, the columns of the matrix represent the UAVs participating in the current bidding task, and D(j,i) represents the distance from the i-th UAV to the j-th target.

[0080] In order to simplify the calculation, the range cost in the task allocation efficiency model of the present invention is expressed by the distance from the UAV to the target.

[0081] S8: Task allocation method based on improved genetic algorithm

[0082] Integer permutation coding is used. For n bidding drones, the chromosome is n segments, each of which corresponds to a drone. 1 A drone means attacking the drone of target 1, then w 2 The drones represent the drones that attack target 2, and then w 3 The first drone represents the drone that hits target 3, and so on. The drones that do not hit the target after that represent the bidders. 1 +w 2 +...+w m The chromosome gene segment indicates that the dominant gene represents the drone that performs the strike mission (the dominant gene is represented by k), which is greater than w 1 +w 2 +...+w m The chromosome gene segment represents a recessive gene (recessive gene is represented by p).

[0083] The encoding method designed by the present invention has the following advantages:

[0084] 1) The integer permutation coding method is used to match the actual swarm drone data link number, and the data link number can represent the gene number in the coding;

[0085] 2) Dominant and recessive genes with distinctive characteristics are designed in the genetic algorithm. The dominant gene corresponds to the successful bidder in the contract network, and the recessive gene corresponds to the bidder in the contract network. From another perspective, the dominant gene corresponds to the drone that performs strike missions, and the recessive gene corresponds to the drone that does not perform any mission. The existence of recessive genes is of great significance, and individuals with better adaptability may appear in the process of genetic variation, which is consistent with the characteristics of biological genetics and the characteristics of drone swarm task allocation.

[0086] S9: After completing the gene encoding, an initial population is randomly generated as the starting solution. Each generation of individuals is processed through a genetic algorithm, the fitness function is calculated, and individuals whose fitness meets the set threshold are selected for the next generation of processing.

[0087] Specifically, the processing flow of the genetic algorithm includes:

[0088] S101: After completing the gene encoding, randomly generate an initial population as the starting solution;

[0089] S102: Calculate the fitness function of individuals in the population;

[0090] S103: Select individuals with high fitness in the population to enter the next generation;

[0091] S104: The parent individuals are divided into groups and crossover operations are performed;

[0092] S105: Parent individuals undergo random mutation operations according to a certain ratio;

[0093] S106: The parent individual performs a reversal operation;

[0094] S107: Generate offspring population;

[0095] S108: Repeat S102 to S107 until the maximum genetic generation G is satisfied max .

[0096] S10: Repeat S9 until the maximum genetic generation G is met. max , will pass through G max The best individual obtained by generation inheritance is decoded to obtain the best task allocation matrix A

[0097] Perform crossover, mutation and reversal on each generation, then bring it into the fitness function for evaluation, select individuals with high fitness, and perform crossover, mutation and reversal operations on the next generation. Perform reverse decoding according to the gene encoding process of S8 to obtain the task allocation matrix A of S6.

[0098] As another aspect of the present invention, the technical solution of the present invention is described in detail in combination with a specific embodiment.

[0099] In this embodiment, the Contract Net Protocol is used to realize the information exchange between the swarm drones. The control method is centralized and distributed. The drones perform cost calculations in a distributed manner, and the auctioneer completes the distribution results in a centralized manner.

[0100] When the UAV (Unmanned Aerial Vehicle) in the swarm drone system is performing a mission and finds that the target drone is "bidding" as the auctioneer, all the swarm drones that receive the bidding information will decide whether to participate in the "bidding" based on their own situation and performance evaluation function. The auctioneer will then decide the "winning bidder" based on the situation of the drones participating in the bidding and the performance evaluation function. The auctioneer will publish the "winning bid" information and the two parties will shake hands to reach an agreement.

[0101] The content of the present invention mainly focuses on how to achieve dynamic task allocation of a swarm drone system as follows.

[0102] A task allocation method for a swarm UAV system based on an improved genetic algorithm comprises the following steps:

[0103] The first step is to quantify the task objectives

[0104] Define the destructive capability of a swarm drone as 1. To attack a target, w is needed. j The equivalent value of the target is w j .

[0105] It is assumed that the anti-damage capabilities of conventional mission targets are known, that is, the equivalent values ​​of different strike targets in this project are known.

[0106] The second step is to allocate strikes based on target equivalence.

[0107] Strike allocation method based on target equivalence: Determine how many drones are assigned to perform the strike mission based on the target equivalence value of the mission. For example, there are three targets, and the target equivalence values ​​are w=3, w=1, and w=2 respectively. The tenderer bids for a total of 6 drones, and seven drones in the swarm unmanned system have submitted bids for the tender information. Based on the performance evaluation function, the tenderer assigns UAV1, UAV4, and UAV6 to strike target 1, assigns UAV3 to strike target 2, and assigns UAV2 and UAV5 to strike target 3. UAV7 did not win the bid for this round of strike missions. The allocation results are as follows: Figure 1 As shown. From the perspective of the mission objective, the quantized value of objective 1 is 3, so the sum of the first row in the allocation principle is 3, and the number of allocated drones cannot be more or less than 3; from the perspective of the drone, the maximum sum of each column is 1, indicating that the drone can only perform one strike mission at a time. When the sum of the numbers in a column corresponding to the drone is 0, it means that the drone has failed to win the bid and becomes a bidder for the mission after the allocation algorithm.

[0108] Step 3: Task allocation effectiveness evaluation model

[0109] The performance evaluation algorithm of the task allocation of the swarm drone system is task-centric, and the swarm drones that discover the task target act as the bidder, that is, as the decision maker of the task allocation. In this embodiment, the task allocation technology of the swarm drone system only considers the path cost PathCost when performing performance evaluation.

[0110] Bidding is conducted for three targets found within a range (the number of mission targets within the range can vary). In the process of allocating strikes, the effectiveness evaluation is the sum of the path costs of all strike drones to the strike targets, and its minimum value is solved.

[0111]

[0112] Among them, w 1 represents the number of drones assigned to attack target 1, w 2 represents the number of drones assigned to attack target 2, w 3 It indicates the number of drones assigned to attack target 3. pu indicates the drone position and pg indicates the target position.

[0113] Step 4: Task allocation method based on improved genetic algorithm

[0114] (1) Solve the path cost matrix D

[0115] It is required to solve the path cost matrix D from the bidding drone to the mission target, which contains the distance between the swarm drone and the mission target, so as to bring it into the performance evaluation function, solve the minimum path cost, and then carry out the strike allocation.

[0116] The path cost between the drone and the target is calculated as follows:

[0117]

[0118] Where D(j,i) represents the distance from the i-th aircraft to the j-th target, pux i ,puy i are the position coordinates of the i-th drone, pgx j , pgy j are the position coordinates of the j-th target respectively.

[0119] Table 1 Algorithm for solving the path cost matrix between bidding drones and targets

[0120]

[0121]

[0122] (2) Improve the genetic algorithm process

[0123] The flowchart of the task allocation method based on genetic algorithm is as follows Figure 2 As shown in the figure, the path cost matrix D of the bidding drone to reach the task target in the task cluster, the position coordinates of the swarm drone system when entering the allocation process, the task target equivalent value w, and the position of the task target pg are input, and finally the optimal strike allocation matrix A is generated through integer coded genetic algorithm optimization.

[0124] (3) Coding

[0125] The integer permutation coding method is adopted. For n bidding drones, the chromosome is n segments, each of which corresponds to a drone. For m targets in a task cluster, the weight of the jth task target is w j There are 6 swarm drones attacking 3 targets in the mission cluster, and the weights of the three targets are w 1 =3, w 2 =2, w 3 = 1. For the task allocation problem of 6 drones {1,2,3,4,5,6} and 3 mission targets, |1|5|2|6|3|4| is a legal chromosome, where |1|5|2| represents the strike mission target g 1 Three drones, |6|3| represents the strike mission target g 2 Two drones, |4| represents the strike mission target g 3 (Each individual has only one chromosome, so chromosome and individual have the same meaning)

[0126] (4) Dominant genes and recessive genes

[0127] For the task allocation with 8 drones {1,2,3,4,5,6,7,8} bidding, there are three task objectives in the task cluster, and their weights are w 1 =3, w 2 =2, w 3 = 1. For a legal chromosome |1|5|2|6|3|4|7|8|, the first six digits |1|5|2|6|3|4| are the drones that won the bid to strike the mission target, that is, the successful bidder, so it is a dominant gene in the individual chromosome (dominant gene is represented by k); the last two digits |7|8| represent the drones that did not win the bid after the bidding, that is, the bidders, so it is a recessive gene in the chromosome (recessive gene is represented by p).

[0128] (5) The specific processing flow of the genetic algorithm includes:

[0129] S101: After completing the gene encoding, randomly generate an initial population as the starting solution

[0130] After completing the chromosome encoding, an initial population must be generated as the starting solution, so the number of the initial population must be determined first. The number of the initial population is generally obtained based on experience. In general, the number of populations depends on the scale of the task target, and its value fluctuates between 10 and 200.

[0131] S102: Calculate the fitness function of individuals in the population

[0132] Assume |k i1 |k i1 |···|k ij |···|k im |p|p|p| is a chromosome encoded with integers, k represents a dominant gene, p represents a recessive gene, is the distance from the i-th UAV to the j-th mission target, and substitutes the elements of the corresponding path cost matrix D. When solving the fitness, the influence of recessive genes is not considered. Then the individual is |k i1 |k i1 |···|k ij |···|k im |Fitness is calculated as:

[0133]

[0134] Among them, s represents the sth individual in the population, f(s) represents the fitness of the sth individual, and t represents the sum of all target equivalent values.

[0135] That is, the fitness function is the inverse of the sum of the distances of the bidding drones from their respective targets after they are assigned to strike targets. Solving for the minimum path cost is equivalent to solving for the maximum individual fitness f. The goal of optimization is to select individuals with the largest fitness value possible. Individuals with larger fitness values ​​are of higher quality, and vice versa. t is equal to the sum of the equivalent values ​​of all mission targets. Converting the solution of maximum efficiency into the solution of maximum fitness is more in line with people's understanding of genetics.

[0136] When considering different scenarios or risk costs, the fitness calculation formula is as follows:

[0137]

[0138] Among them, U represents the task allocation efficiency.

[0139] S103: Select individuals with high fitness in the population to enter the next generation

[0140] The selection operation is to select individuals from the parent population to the new population with a certain probability. The probability of an individual being selected is related to the fitness value. The larger the fitness value, the greater the probability of being selected. The fitness ranking method is adopted to sort the top P individuals with the highest fitness. s % of individuals are selected to enter the offspring, and subsequent operations such as crossover and mutation are performed, and the best individuals in each generation are retained.

[0141] S104: Parent individuals are grouped in pairs and crossover operations are performed

[0142] Use partial mapping crossover to determine the parent generation of the crossover operation, group the parent generation samples in pairs, and repeat the process for each group (assuming that the number of task clusters is 5). c is the crossover probability, usually selected as 0.4 to 0.99:

[0143] (1) Generate two random integers r in the interval [1,n] 1 and r 2 (n represents all the drones participating in the bidding, and n is 8 this time), determine two positions, and cross the data between the two positions, such as r 1 =2, r 2 =4 Because the starting point of the swarm drone system is not fixed and can only be ranked at the first coding point, the crossover starts from the second coding point.

[0144] 13|46|2587

[0145] 12|53|4687

[0146] After the cross

[0147] 1*|53|2*87

[0148] 12|46|**87

[0149] (2) After crossover, if there are repeated code numbers in the same individual, the non-repeated numbers are retained, and the conflicting numbers (with *) are eliminated by partial mapping, that is, mapping is performed using the corresponding relationship of the middle segment. The result is:

[0150] 14|53|2687

[0151] 12|46|3587

[0152] Deconfliction prevents the same number from appearing multiple times and the same drone from performing multiple strike missions, resulting in a situation that is inconsistent with the actual swarm of drones.

[0153] S105: Parent individuals undergo random mutation operations according to a certain ratio

[0154] Mutation probability P m Generally, the probability of all genes of an individual mutating is generally 0.001 to 0.1. The mutation strategy is to randomly select two genes and swap their positions. Generate two random integers r in the range [1, n] 1 and r 2 (n represents all the drones participating in the bidding, and n is 8 in this case), determine two positions and swap them, for example r 1 =2, r 2 =4.

[0155] 13|46|2587

[0156] After mutation:

[0157] 16|43|2587

[0158] By exchanging positions, no new data link numbers will be generated outside the bidding drones, which is in line with the actual swarm situation, and there will be no duplication of numbers, thus avoiding the contradiction of the same drone performing multiple strike missions.

[0159] S106: Reversal operation of parent individuals

[0160] Randomly generate two random integers r in the range [1,n] 1 and r 2 (n represents all the drones participating in the bidding, and n is 8 in this case), determine two positions and reverse their positions, for example, r 1 =2, r 2 =6.

[0161] 14|5326|87

[0162] After reversal:

[0163] 14|6235|87

[0164] In order to improve the local search capability of the genetic algorithm, multiple evolutionary reversal operations are performed continuously after selection, crossover, and mutation. Here, "evolution" refers to the unidirectional nature of the genetic operator, that is, only those with improved fitness values ​​after reversal are accepted, otherwise the reversal is invalid.

[0165] S107: Generate offspring population

[0166] The parent population generates a new offspring population through fitness calculation, selection, crossover, mutation, and reversal.

[0167] S108: Repeat S102 to S107 until the maximum genetic generation G is satisfied max , will pass through G max The best individual obtained by generation inheritance is decoded to obtain the best task allocation matrix A

[0168] Perform crossover mutation on each individual, then bring it into the fitness function for evaluation, select individuals with large fitness, and perform crossover mutation and reversal operations on the next generation.

[0169] Repeat the operation until the maximum genetic generation G is met. max .

[0170] By G max Through generations of genetic variation, the optimal or near-optimal individuals for task allocation have been found.

[0171] |1|5|2|6|3|4|7|8|

[0172] According to the equivalent value of each task target, the above optimal allocation individuals are decoded to obtain the task allocation matrix A.

[0173] (6) Verification of task allocation effect

[0174] A preliminary simulation verification was conducted for the task allocation method based on the improved genetic algorithm. Six UAVs were set up in the simulation, and their positions are shown in Table 2 below; three mission targets need to be attacked, and their position information and equivalent values ​​are shown in Table 3 below.

[0175] Table 2 Mission target information

[0176] Target Number pux puy 1 106.40 66.74 2 105.14 69.22 3 104.99 71.49 4 104.73 73.14 5 106.37 66.65 6 107.05 64.11

[0177] Table 3 Mission target information

[0178]

[0179]

[0180] The attack distribution matrix A obtained by the simulated distribution algorithm of the improved genetic algorithm is as follows:

[0181]

[0182] It can be seen from the task allocation matrix A that UAV1 is assigned to attack target 3, UAV2, UAV3, and UAV4 are assigned to attack target 1, and UAV5 and UAV6 are assigned to attack target 2.

[0183] Table 4 Task allocation capability based on improved genetic algorithm

[0184] Dynamic task allocation capability Improved genetic algorithm Bidding drone number is uncertain √ Drone location is uncertain √ Uncertain number of drones bidding √ Uncertain number of mission targets √ Mission target equivalent value is uncertain √

[0185] Table 5 Simulation time using improved genetic algorithm

[0186] Serial number Genetic Algorithms Serial number Genetic Algorithms 1 0.403 6 0.548 2 0.524 7 0.551 3 0.547 8 0.557 4 0.550 9 0.552 5 0.551 10 0.539

[0187] The task allocation method based on improved genetics can quickly find a relatively optimal allocation strategy, and the dynamic task allocation method based on improved genetic algorithms can complete dynamic task allocation when the bidding drone number, bidding drone location, task target number, and task target equivalent value are uncertain, and has high dynamic adaptability. However, the dynamic task allocation method based on improved genetic algorithms cannot guarantee that the optimal solution can be found every time, but can only guarantee that the suboptimal or optimal solution can be found within the range that meets the requirements, and the probability of finding the optimal solution decreases as the number of bidding drones and the number of task targets increase.

[0188] Features described and / or illustrated above for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or used in place of features in other embodiments.

[0189] It should be emphasized that the term "include / comprises" when used herein refers to the presence of features, integers, steps or components, but does not exclude the presence or addition of one or more other features, integers, steps, components or combinations thereof.

[0190] The many features and advantages of these embodiments are apparent from this detailed description, and thus the appended claims are intended to cover all such features and advantages of these embodiments that fall within their true spirit and scope. Furthermore, since numerous modifications and changes will readily occur to those skilled in the art, it is not intended that the embodiments of the invention be limited to the exact construction and operation illustrated and described, but rather all suitable modifications and equivalents falling within the scope thereof are intended to be covered.

[0191] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0192] Parts of the present invention that are not described in detail are well known to those skilled in the art.

Claims

1. A task allocation method for swarm drone system based on improved genetic algorithm. It is characterized in that The steps include: Establish a mission objective library of typical targets in the combat area; Quantify the destructive power of the swarm drones used, and set the destructive power equivalent value of a single drone to 1; According to the damage capability equivalent value of a single UAV, determine the anti-damage capability equivalent value of each target in the mission target library; The task allocation efficiency evaluation model is constructed based on the difference between the total target value benefit and the total cost; Construct a task allocation matrix with the number of targets as rows and the number of drones as columns; The drone that finds the target during the flight of the swarm drone acts as the auctioneer and conducts bidding; The bidding drones will use integer arrangement coding to encode chromosome genes according to the data link number sequence; An initial population is randomly generated for genetic algorithm processing, the fitness function is calculated for each generation of individuals, and individuals whose fitness meets the set threshold are selected for the next generation of genetic processing; Until the maximum genetic generation is met, the best individual is decoded to obtain the optimal task allocation matrix.

2. The task allocation method according to claim 1, It is characterized in that The anti-damage capability equivalent value of the target is determined through combat examples, actual combat exercises or simulation verification methods.

3. The task allocation method according to claim 1, It is characterized in that The swarm of drones uses a contract network protocol to exchange information and adopts a distributed control system. Each drone performs cost calculation in a distributed manner, and the auctioneer completes and sends the allocation results in a centralized manner.

4. The task allocation method according to claim 1, It is characterized in that The chromosome includes a dominant gene and a recessive gene, wherein the dominant gene indicates a UAV that performs a strike mission, and the recessive gene indicates a UAV that does not perform a strike mission.

5. The task allocation method according to claim 1, It is characterized in that The task allocation efficiency evaluation model is: U=Reward-Path Cost-Risk Cost Where j represents the target number, m represents the total number of targets in the area; g j represents the jth target, represents the i-th drone that attacks target j, represents the position of UAV i that is executing the attack on target j, pg j represents the position of target j; w j represents the equivalent value of the jth target; q j represents the profit coefficient of the jth target; β j represents the risk factor of the jth target; Reward is the value of the target, Reward(g j ) represents the value of target j, Path Cost is the cost of the journey, Path Cost (pu i ,pg j ) represents the flight cost of UAV i to target j; Risk Cost is the risk cost, Risk Cost (pu i ,pg j ) represents the risk of UAV i being destroyed on its way to target j; U represents the task allocation efficiency.

6. The task allocation method according to claim 5, It is characterized in that The fitness function is: Among them, U represents the task allocation efficiency, s represents the sth individual in the population, and f(s) represents the fitness of the sth individual.

7. The task allocation method according to claim 1, It is characterized in that The task allocation efficiency evaluation model is: Among them, j represents the target number, and m represents the total number of targets in the area; represents the position of UAV i that is executing the attack on target j, pg j represents the position of target j; w j represents the equivalent value of the jth target, Path Cost (pu i ,pg j ) represents the flight cost from UAV i to target j; U represents the task allocation efficiency.

8. The task allocation method according to claim 7, It is characterized in that The fitness function is: Among them, s represents the sth individual in the population, f(s) represents the fitness of the sth individual, and t represents the sum of all target equivalent values. is the distance from the i-th UAV to the j-th mission target.

9. The task allocation method according to any one of claims 1 to 8, It is characterized in that The genetic manipulation includes Select individuals with high fitness in the population to enter the next generation; The parent individuals are grouped in pairs for crossover operation; Parent individuals undergo random mutation operations according to a certain ratio; The parent individual performs a reversal operation.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that The processor executes the computer program to implement the task allocation method described in any one of claims 1 to 8.

Images