Unmanned aerial vehicle cluster collaborative rescue task allocation method based on population interactive learning type particle swarm algorithm

Abstract

The invention discloses an unmanned aerial vehicle cluster collaborative rescue task allocation method based on a population interactive learning type particle swarm algorithm, and the method comprises the following steps: determining a minimum rescue comprehensive penalty objective function based on the rescue value of each rescue task and the time of each unmanned aerial vehicle from a base to the responsible rescue task; determining a shortest rescue time objective function based on the time when each unmanned aerial vehicle arrives at the last responsible rescue task; according to the minimum rescue comprehensive penalty objective function and the shortest rescue time objective function, constructing an unmanned aerial vehicle cluster collaborative rescue task allocation model; and solving the unmanned aerial vehicle cluster collaborative rescue task allocation model based on a population interactive learning type particle swarm algorithm, and allocating unmanned aerial vehicle cluster collaborative rescue tasks based on a solving result. The method can effectively solve the problem of cooperative multi-task allocation of the unmanned aerial vehicle cluster so as to improve the efficiency of post-disaster rescue work.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle task allocation, and more specifically to a method for allocating unmanned aerial vehicle cluster collaborative rescue tasks based on a population interactive learning particle swarm algorithm. Background Art

[0002] In the traditional post-disaster rescue process, there are often challenges such as insufficient information and safety risks. The introduction of UAVs (Unmanned Aerial Vehicles) provides a new possibility for solving rescue work. The key to post-disaster rescue is to deliver relief supplies to as many trapped people as possible in the shortest time. Therefore, according to the pre-obtained disaster and UAV information, accurately planning the tasks and execution order of each UAV, that is, task allocation, has become the core link of rescue material delivery. Therefore, reasonable task allocation is the core to ensure that UAVs can efficiently perform various tasks. For the problem of multi-UAV task allocation, the above process includes two stages: allocation and sorting. Allocation refers to assigning multiple tasks to different UAVs; sorting refers to sorting the execution order of the assigned tasks.

[0003] It is easy to know that after a disaster, the mission information is uncertain. Due to the interference caused by the collapse of buildings, rescuers often cannot accurately obtain the number of trapped people, and it is difficult to carry out rescue operations effectively. Therefore, the study of UAV task allocation strategy in uncertain environments is of great significance.

[0004] In recent years, scholars have conducted extensive research on task allocation problems for different scenarios and types of drones, and proposed a variety of drone task allocation models and corresponding algorithms suitable for different problems and scenarios. However, most existing methods are based on precise target design. For fuzzy and uncertain task situations, due to the addition of uncertainty factors, in scenarios such as multi-target, multi-preference and strong constraints, traditional task allocation methods are slightly insufficient in effectiveness.

[0005] Particle Swarm Optimization (PSO) has performed well in solving the problem of coordinated task allocation of UAVs in recent years due to its small number of parameters, insensitivity to parameters, and easy expansion, and there have been many successful application cases. Therefore, in an environment with uncertain task information, especially when faced with rescue tasks with different weights and time sensitivity, how to use the particle swarm algorithm to effectively solve the problem of coordinated multi-task allocation of UAV clusters to improve the efficiency of post-disaster rescue work is an urgent problem that technicians in this field need to solve. Summary of the invention

[0006] In view of the above problems, the present invention provides a method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm to at least solve some of the technical problems mentioned in the above background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] The present invention provides a method for allocating unmanned aerial vehicle cluster collaborative rescue tasks based on a population interactive learning particle swarm algorithm, comprising the following steps:

[0009] Based on the rescue value of each rescue mission and the time it takes for each drone to arrive at the rescue mission it is responsible for from the base, determine the minimum rescue comprehensive penalty objective function;

[0010] Determine the shortest rescue time objective function based on the time it takes each drone to reach the last rescue mission it is responsible for;

[0011] According to the minimum rescue comprehensive penalty objective function and the shortest rescue time objective function, a UAV cluster collaborative rescue task allocation model is constructed;

[0012] The UAV cluster collaborative rescue task allocation model is solved based on the population interactive learning particle swarm algorithm, and the UAV cluster collaborative rescue tasks are allocated based on the solution results.

[0013] Furthermore, the minimum rescue comprehensive penalty objective function is expressed as:

[0014]

[0015] Among them, f1 represents the minimum rescue comprehensive penalty objective function; I represents a total of I drones; i represents the i-th drone; J represents a total of J rescue missions; j represents the j-th rescue mission; represents the time it takes for drone i to arrive at rescue mission j from the base; Represents the rescue value of rescue mission j at time t.

[0016] Furthermore, the time when each drone arrives at the rescue mission from the base, and the corresponding acquisition steps include:

[0017] Based on the maximum empty-load flight speed of the drone, the maximum cargo capacity of the drone, and the amount of rescue materials each drone carries before arriving at each rescue mission, the flight speed of each drone between every two rescue missions is obtained; it is expressed as:

[0018]

[0019] Among them, vel i,jrepresents the flight speed of UAV i between rescue mission j-1 and rescue mission j; rescue mission j-1 is the previous rescue mission of rescue mission j; ε represents the UAV performance impact coefficient; vel max Indicates the maximum flight speed of the drone without load; S max Indicates the maximum cargo capacity of the drone; S i,j represents the amount of rescue supplies of UAV i before it arrives at rescue mission j;

[0020] Based on the flight speed and distance of each drone between each rescue mission, the time it takes for each drone to arrive at the rescue mission it is responsible for from the base is obtained, expressed as:

[0021]

[0022] in, represents the time it takes for UAV i to arrive at rescue mission j from the base; J i represents the number of rescue missions performed by UAV i from its departure from the base to its return to the base, and represents the distance between rescue mission j-1 and rescue mission j; ΔT represents the sum of the unloading time and replenishment time of drone i.

[0023] Furthermore, the shortest rescue time objective function is expressed as:

[0024]

[0025] Among them, f2 represents the shortest rescue time objective function; represents the time when UAV i arrives at the last rescue mission it is responsible for; end represents the last rescue mission; I represents a total of I UAVs; ΔT represents the sum of the unloading time and the replenishment time of UAV i.

[0026] Furthermore, the encoding and decoding strategies of the population interactive learning particle swarm algorithm specifically include:

[0027] In the encoding stage: integer encoding is used, and the particle structure is represented by a two-stage method; the first stage is a nested array, whose size is a 1*I matrix, indicating that there are I inner arrays, and the i-th column array X 1i The inner size is 1*n i , represents the rescue mission assigned to UAV i and its execution order, n i is the number of rescue missions assigned to UAV i; suppose there are J rescue missions in total, so X 1i The value range is [1,J]; the second stage is 1*ns i The array X 2i , represents the resupply time of UAV i, which is represented by the rescue mission number, that is, the UAV will be resupplied after completing the mission, nsi is the number of times UAV i needs to be resupplied during its mission, and its value range is {1,2,…,n i -1};

[0028] In the decoding stage: a two-stage embedding method is used, that is, in the first stage, the starting point, the replenishment point, and the end point are embedded according to the array content in the second stage; if the i-th particle X i To decode, first X i The jth column of the first stage is decoded to separate the task allocation of the i-th UAV, that is, a series of task points, represented by X i ; According to the supply point location information given by the corresponding position in the second stage, the supply point is embedded in the task; the embedding operation follows the principle of proximity, that is, the supply point closest to the first rescue task to be performed after the drone is resupplied is selected, and finally the base is embedded at both ends of the drone task allocation sequence, thereby obtaining the decoded drone's real flight path, which includes the starting point, rescue point, supply point and end point.

[0029] Furthermore, the particle swarm algorithm based on population interactive learning is used to solve the drone cluster collaborative rescue task allocation model, specifically including:

[0030] Initialize the population interactive learning particle swarm algorithm based on K-means clustering to obtain the initial population;

[0031] Divide the initial population into a main population and an auxiliary population;

[0032] Determine the particles to be updated in the main population, and select particle learning objects based on the auxiliary population;

[0033] Based on the particle learning object, the particles to be updated are updated.

[0034] Furthermore, the K-means clustering-based population interactive learning particle swarm algorithm is initialized to obtain an initial population, specifically including:

[0035] (1) Based on the location information of J rescue tasks, the K-means clustering algorithm is used to divide the J rescue tasks into I sub-clusters; I sub-clusters are randomly assigned to I drones, thereby completing the initial task assignment;

[0036] (2) Based on the comprehensive benefits corresponding to the rescue missions that each drone is responsible for, the rescue missions of each drone are ranked; specifically: let rescue mission j be one of the rescue missions that drone i is responsible for, and the comprehensive benefit of selecting rescue mission j as the next rescue mission to be performed by drone i is:

[0037]

[0038] in, represents the comprehensive benefit of drone i choosing rescue mission j as the next rescue mission at time t; v j (t) represents the rescue value of rescue mission j; represents the time it takes for drone i to arrive at rescue mission j from the base; d i,j represents the distance between rescue mission j and drone i; α represents the weight;

[0039] (3) Under the current rescue mission execution order of drone i, resupply the drone; specifically: suppose that after drone i completes the jth rescue mission assigned to it, the total amount of rescue supplies unloaded by drone i is At this time, the supply probability of drone i is expressed as:

[0040]

[0041] in, represents the probability of resupply of UAV i after completing the jth rescue mission; S max represents the maximum cargo capacity of the drone; in particular, when and , it means that UAV i cannot complete the j-th rescue mission assigned to it. In this case, after completing the j-1-th rescue mission, the UAV must be resupplied.

[0042] Furthermore, the dividing the initial population into a main population and an auxiliary population specifically includes:

[0043] Perform non-dominated sorting in the form of interval numbers on the initial population Pop to obtain a new population Pop' with new particle sorting;

[0044] According to the principle of the least common multiple of integers, determine the smallest integer θ divided by the ratio θ int ;

[0045] According to the non-dominated sorting result, select the first θ from the new population Pop' int Particles constitute a minimum cycle unit, and in the minimum cycle unit, particles are distributed to the main population Pop1 and the auxiliary population Pop2 according to the proportion θ; continue to select the next θ according to the non-dominated sorting result int The particles are divided and the process is repeated until all particles in the new population Pop' are allocated.

[0046] Furthermore, the determining of the particle learning object based on the main population and the auxiliary population specifically includes:

[0047] The particles in the auxiliary population that are not dominated by the particles to be updated are taken as the range of particle learning objects;

[0048] The particle with the shortest vertical distance to the line connecting the particle to be updated and the origin is selected as the particle learning object within the range.

[0049] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm, which has the following beneficial effects:

[0050] The present invention takes the minimum rescue penalty and the minimum task completion time as evaluation indicators, and establishes a UAV cluster collaborative rescue task allocation model under uncertain environment. This model can effectively solve the problem of UAV cluster collaborative multi-task allocation to improve the efficiency of post-disaster rescue work.

[0051] The population interactive learning particle swarm algorithm provided by the present invention significantly accelerates the population convergence speed and optimizes the distribution of the Pareto frontier through parallel updating of the main and auxiliary populations, information interaction and collaborative evolution. The algorithm can not only stably solve the problem of cooperative task allocation of drone clusters, but also has good robustness and stability, and shows better adaptability in complex scenarios.

[0052] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0054] Figure 1 A schematic flow chart of a method for allocating rescue tasks among drone clusters based on a population interactive learning particle swarm algorithm provided in an embodiment of the present invention.

[0055] Figure 2 A schematic diagram of rescue mission allocation provided in an embodiment of the present invention.

[0056] Figure 3 A schematic diagram of a population interactive learning particle swarm algorithm framework provided in an embodiment of the present invention.

[0057] Figure 4 A schematic diagram of initialization task allocation provided by an embodiment of the present invention.

[0058] Figure 5 A schematic diagram of population division provided in an embodiment of the present invention.

[0059] Figure 6A schematic diagram of learning object selection provided in an embodiment of the present invention.

[0060] Figure 7 A schematic diagram of a learning object selection proof provided in an embodiment of the present invention.

[0061] Figure 8 A schematic diagram of a two-stage update provided by an embodiment of the present invention.

[0062] Fig. 9 A schematic diagram of updating the internal population of a learning library provided in an embodiment of the present invention.

[0063] Fig.10 A schematic diagram of evaluation indicators provided by an embodiment of the present invention.

[0064] FIG11( a ) is a schematic diagram of the box shapes obtained by the PILPSO provided by an embodiment of the present invention and four comparison algorithms in scenario 1. FIG.

[0065] FIG11( b ) is a schematic diagram of the box shapes obtained by the PILPSO provided by an embodiment of the present invention and four comparison algorithms in scenario 2. FIG.

[0066] FIG11( c ) is a schematic diagram of the box shapes obtained by the PILPSO provided by an embodiment of the present invention and four comparison algorithms in scenario 3. FIG.

[0067] FIG12( a ) is a schematic diagram of the Pareto front distribution of the five algorithms provided in an embodiment of the present invention in scenario 1. FIG.

[0068] FIG12( b ) is a schematic diagram of the Pareto front distribution of the five algorithms provided in an embodiment of the present invention in scenario 2.

[0069] FIG12( c ) is a schematic diagram of the Pareto front distribution of the five algorithms provided in an embodiment of the present invention under scenario 3. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] See also Figure 1 As shown, the embodiment of the present invention discloses a method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm, comprising the following steps:

[0072] Based on the rescue value of each rescue mission and the time it takes for each drone to arrive at the rescue mission it is responsible for from the base, determine the minimum rescue comprehensive penalty objective function;

[0073] Determine the shortest rescue time objective function based on the time it takes each drone to reach the last rescue mission it is responsible for;

[0074] According to the minimum rescue comprehensive penalty objective function and the shortest rescue time objective function, a UAV swarm collaborative rescue task allocation model is constructed;

[0075] The UAV swarm collaborative rescue task allocation model is solved based on the population interactive learning particle swarm algorithm, and the UAV swarm collaborative rescue tasks are allocated based on the solution results.

[0076] Next, the above contents will be described in detail.

[0077] 1. Problem description and model establishment

[0078] 1.1. Problem Description

[0079] In order to simplify the problem and maintain generality, the following assumptions are made:

[0080] 1) When the UAV performs a mission with a fixed load, the flight speed is constant;

[0081] 2) Use drones to detect vital signs in the environment before rescue and determine the location of each rescue point;

[0082] 3) During the entire mission execution, the positions of trapped persons at all rescue points remain unchanged;

[0083] 4) The material supply station (i.e. supply point) has sufficient resources to ensure the completion of the mission.

[0084] Based on the above assumptions, in a post-disaster environment, suppose there are I drones parked at the same base, and each drone i corresponds to a rescue mission list O i ; m supply points, scattered at several accessible fixed points in the area, and their positions are known; there are J rescue points in total, which constitute the rescue task set [1,J], and the positions of the rescue points are known, expressed as P = {(x1,y1),(x2,y2),…,(x J ,y J )}. UAVs have certain limitations in range, cargo capacity, and speed, requiring cooperation between UAVs. Under the condition of uncertain mission information, the rescue value information at different times is calculated, and the mission execution sequence is planned for each UAV accordingly, so as to complete the mission in the shortest time under the premise of minimum rescue penalty.

[0085] The schematic diagram of multi-UAV coordinated rescue can be seen in Figure 2As shown, there are 3 drones in the environment; 12 missions require rescue, each rescue mission corresponds to a rescue point; and two supply points. Figure 2 In the figure, arrows of different colors indicate the rescue paths of different aircraft. Taking the rescue path of UAV 1 as an example, the rescue order of its mission is rescue point 1→supply station→rescue point 8→rescue point 2→rescue point 4.

[0086] 1.2 Calculation of rescue value

[0087] In order to complete the task in the shortest time with the minimum rescue penalty, it is necessary to judge the rescue value of each task according to its specific conditions. The rescue value is usually related to the number of rescuers, the difficulty of rescue and the time of rescue, but the number of rescuers can more directly judge the value of a rescue point. After the disaster, due to the limited high-altitude reconnaissance information, the approximate number of trapped people can only be judged by the scale of the building, the degree of damage and other pre-mastered information. Therefore, the embodiment of the present invention uses the building damage map obtained by satellite or high-altitude drone, based on the model and results constructed by the "Earthquake Casualty Assessment Method and Application Based on Building Structural Destruction" published in the "Journal of Tsinghua University (Natural Science Edition)" in 2015 (hereinafter referred to as Document 1), to judge and infer the damage to the building where the rescue point is located. The present invention calculates the number of trapped people in the rescue point based on this result.

[0088] In the embodiment of the present invention, the rescue priority is divided into three levels: severe (H), moderate (M) and light (L) injuries, and the number of wounded at each rescue point is determined based on prior information. Taking into account the complexity and uncertainty of the post-disaster environment, the number of people calculated by Document 1 is not accurate enough. Therefore, the embodiment of the present invention uses interval numbers to represent the number of wounded at each level. Specifically, assuming that the calculation result based on the model of Document 1 is Pr, the interval number of people in the present invention is expressed as [Pr(1-err), Pr(1+err)], where err is the ratio of the error mean value between the calculation result of the model of Document 1 and the actual result. It can be seen from the results of Document 1 that the err of the number of lightly injured is 27%, the err of the number of moderately injured is 20%, and the err of the number of seriously injured is 12%. Then, the example of the present invention determines the interval of the number of people for each level of the wounded accordingly.

[0089] When setting the function, the sum of the intervals of the number of injured persons with a single degree of injury in the entire rescue area is taken as the benchmark, and the ratio of the intervals of the number of injured persons with different degrees at different rescue points to the total is used as the calculation basis. The absolute number is converted into a relative number to make it more in line with the actual situation, as shown in formula (1):

[0090]

[0091] in, as well as They represent the number of seriously injured, moderately injured and slightly injured people in rescue point j respectively; the total number of seriously injured people is Slander Minor injuries

[0092] During the rescue process, the condition of the injured will deteriorate over time, especially for moderately and severely injured people, so the rescue value of the rescue point will also increase over time. and Part will grow, which can be calculated by formula (2-1) and formula (2-2):

[0093]

[0094]

[0095] Among them, γ1 and γ2 represent the growth coefficients, and their values ​​are set according to the actual situation. t is the time, and the final rescue value is is a 3×2 matrix; v h (t), v m (t) and v l (t) represents the rescue value of severe injury, moderate injury and light injury respectively; the rescue value weights of each injury level from severe to light are ω1, ω2 and ω3, and the weight values ​​are determined by the decision maker according to the actual situation; and Same reason.

[0096] 1.3 Uncertain Optimization Model

[0097] The embodiment of the present invention considers two objective functions: the minimum comprehensive rescue penalty objective function f1 and the shortest rescue time objective function f2. In f1, priority is given to tasks with higher rescue value, and the integral value of the growth curve of its rescue value over time is used as the rescue penalty. As described in Section 1.2, the rescue value is an interval number, so f1 is constructed as an interval function. The other objective function f2 aims to complete all rescue tasks in the shortest time. Since these two objectives conflict, the embodiment of the present invention adopts two objective functions to provide decision makers with different options for selection.

[0098] In model construction, the impact of the amount of relief supplies on the flight speed of the drone also needs to be considered. In addition to the drone base, multiple resource supply points are set up in the area for drones to replenish supplies. Therefore, the decision variables of the model constructed in the embodiment of the present invention include task allocation, task scheduling after allocation, and the selection of replenishment time.

[0099] 1.3.1 Objective Function

[0100] (1) Minimum rescue comprehensive penalty objective function f1, expressed as:

[0101]

[0102] Among them, f1 represents the minimum rescue comprehensive penalty objective function; I represents a total of I drones; i represents the i-th drone; J represents a total of J rescue missions; j represents the j-th rescue mission; represents the time it takes for drone i to arrive at rescue mission j from the base; represents the rescue value of rescue mission j at time t (the rescue values ​​of the base and the supply station are both 0), which is calculated by formula (2) and is a 3×2 matrix; It represents the cumulative rescue penalty of trapped persons with different degrees of injury when UAV i arrives at rescue mission j, which is also a 3×2 matrix; It represents the time it takes for UAV i to reach rescue mission j from the base; its calculation method is shown in formula (4).

[0103]

[0104] in, represents the time it takes for UAV i to arrive at rescue mission j from the base; J i represents the number of rescue missions performed by UAV i from its departure from the base to its return to the base, and represents the distance between rescue mission j-1 and rescue mission j; ΔT represents the sum of the unloading time and replenishment time of drone i; vel i,j It represents the flight speed of UAV i between rescue mission j-1 and rescue mission j. Its value is related to its cargo capacity. The specific expression is shown in formula (5).

[0105]

[0106] Among them, vel max Indicates the maximum flight speed of the drone without load; S max Indicates the maximum cargo capacity of the drone; S i,j Indicates that drone i has reached rescue mission R ,j The amount of rescue supplies before delivery; ε represents the UAV performance impact coefficient, and its value depends on the UAV performance.

[0107] (2) Shortest rescue time objective function f2

[0108] The rescue time f2 of the drone to complete all rescue tasks can be calculated by formula (6):

[0109]

[0110] in, Arrive at the task execution list O for drone i i end represents the last rescue mission; ΔT represents the sum of the unloading time and replenishment time of UAV i.

[0111] 1.3.2 Constraints

[0112] The drone needs to consider multiple constraints when performing tasks, including battery and load constraints. The specific requirements are as follows:

[0113] (1) The power consumption constraint of the UAV route, that is, the power consumption of each UAV from the base to the base must be less than or equal to the battery capacity of the UAV; it can be expressed as:

[0114]

[0115] Among them, J i represents the number of rescue missions performed by the i-th UAV from its departure from the base to its return to the base. b represents the battery capacity of the drone; ρ represents the distance that can be flown per unit of power;

[0116] (2) UAV load constraint, that is, the total amount of materials unloaded by each UAV after departing from the current base or supply point must be less than or equal to the maximum cargo capacity of the UAV; expressed as:

[0117]

[0118] Among them, O i represents the rescue mission table corresponding to UAV i; J i ' indicates O i The rescue missions carried out consecutively by any two bases / supply stations; i =[0,1,5,1002,7,8,0], then J i '=[1,5]orJ i '=[7,8], where 102 means return to the supply station, and 0 means the base station; Indicates J i 'The corresponding amount of relief supplies; S max Indicates the maximum cargo capacity of the drone;

[0119] (3) Task allocation constraint, that is, each rescue mission can only be assigned to one UAV, and each rescue mission must be assigned to a UAV for execution; expressed as:

[0120]

[0121] in, Represents the sum of the number of all task points in the j-th column of matrix x; represents the number of task j in the i-th column of x, δ() represents the indicator function, when the condition in the brackets is true, δ()=1; otherwise it is 0.

[0122] (4) UAV supply constraint: when a UAV completes a rescue mission, if the remaining rescue supplies do not meet the requirements of the next rescue mission, it needs to return to the nearest supply point to replenish supplies; it can be expressed as:

[0123]

[0124] Among them, O i,j+1 Represents the rescue mission table o corresponding to drone i i The j+1th rescue mission in O i,j Represents the rescue mission table O corresponding to drone i i The j-th rescue mission in ; Indicates that the drone has completed O i,j The amount of relief supplies remaining at the time of the Indicates that the drone has completed O i,j+1 The amount of relief supplies required; M represents an integer, and the value of the present invention is 1000; m represents the number of supply points;

[0125] (5) UAV path constraint, that is, each UAV must start from the base at the beginning and return to the base after completing the rescue mission it is responsible for; it can be expressed as:

[0126] O i,1 =0 (11)

[0127] O i,end =0 (12)

[0128] Among them, O i,1 Indicates the first rescue mission that drone i is responsible for; O i,end Indicates the last rescue mission that drone i is responsible for;

[0129] (6) UAV task quantity constraint, that is, each UAV performs at least one task. It can be expressed as:

[0130]

[0131] Among them, J i It represents the number of rescue missions performed by the i-th UAV from its departure from the base to its return to the base.

[0132] 1.4 Model Conversion

[0133] The value of the minimum rescue comprehensive penalty objective function f1 in the model built in the previous section is an interval number. For the convenience of calculation, this section converts the interval objective function into an exact objective function. In the conversion process, in addition to considering the deterministic information of the interval number, its uncertainty information must also be considered. Therefore, this section will consider both the interval weighted mean and the interval uncertainty degree to convert the interval target value into an exact value. Specifically, by setting the decision maker tolerance coefficient ζ∈[0,1], the minimum rescue comprehensive penalty objective function f1 is converted into a deterministic target, as shown in formula (14).

[0134]

[0135] Among them, Punish u It is used to evaluate the uncertainty of the interval minimum rescue comprehensive penalty objective function f1, and its value is Punish + -Punish - ; Punish m Punish m =Punish - +Punish u ×ζ. If the decision maker is more optimistic, the coefficient ζ can be smaller; otherwise, the value of ζ can be increased. u and Punish m The positions of the original f1 after transformation are determined for the ordinate and abscissa coordinates on the plane respectively, and the position information of the interval number on the plane is converted into the vector length Punish based on formula (14) exact , replacing the fuzzy interval information with precise vector length information.

[0136] In addition, in the algorithm proposed by the present invention, in order to reduce the information distortion caused by the conversion, the objective function value represented by the interval number is used for the comparison of the main population and the non-dominated sorting operation. Therefore, an interval comparison method is defined: weighted two-end subtraction comparison method: let the interval number a = [a - ,a + ], the number of intervals b = [b - ,b + ]; when ζ×(b + -a + )+(1-ζ)×(b - -a - )>0, then b>a, otherwise a>b; if ζ×(b + -a + )+(1-ζ)×(b - -a - )=0, then b=a.

[0137] 2. Algorithm Design

[0138] In order to efficiently solve the above model, the present invention proposes a population interactive learning particle swarm algorithm. Figure 3 The basic process framework of the proposed algorithm is given. First, the population is initialized according to the characteristics of the problem, and the population is non-dominated sorted. Then, according to the given ratio θ and the population sorting result, the population is divided into the main population and the auxiliary population. Among them, the main population is responsible for the iteration and convergence of the algorithm, and is assisted by the distribution of local updates to optimize the solution; the auxiliary population, as an evolvable external archive set, provides guidance for the main population. The two populations run in parallel and exchange some individuals regularly. The two populations use different strategies to execute the basic process of PSO to update individual and global extreme values; at the same time, the main population performs non-dominated sorting of the population in the form of interval numbers at the end of each generation to reduce the impact of information distortion caused by numerical conversion on the reliability of the solution.

[0139] Specifically, in the "problem-guided population initialization strategy" stage, according to the problem characteristics, the strategy is gradually selected to initialize it, and a set of better initial population distribution is obtained; in the "update learning based on population interaction" stage, the auxiliary population is constructed to guide the iterative update of the main population, focusing on accelerating the convergence of the algorithm and optimizing the diversity of the Pareto frontier. The present invention constructs a learning library with the auxiliary population, which is updated in parallel with the main population and serves as the learning and interaction object of the main task.

[0140] Table 1 below gives the pseudo code of the population interactive learning particle swarm algorithm proposed in the present invention. Line 1 executes the population initialization strategy proposed in Section 2.2, and initializes the allocation, sorting, and replenishment time selection in turn; Line 2 executes the population splitting operation before the iterative update operation, and divides the population Pop into the main population Pop1 and the auxiliary population Pop2 according to the given ratio θ and the population sorting result; Lines 3-11 are the iterative update stage, of which Lines 5-6 are the internal self-update stage of Pop2; Line 7 is the selection, update, and local optimization of Pop1 according to the frontier sparse area; Lines 8-9 exchange Pop1 and Pop2 individuals according to conditions; Line 10 is the subsequent processing operation of each generation.

[0141] Algorithm Pseudocode of the proposed algorithm.

[0142] Table 1 The overall operation process of the group interactive learning particle swarm algorithm

[0143] enter Population size Num, maximum number of iterations G Output Optimal solution (1) Initialize PSO parameters, determine task allocation, task sequencing, and replenishment time (2) <![CDATA[Sort the population Pop and split it into the main population Pop1 and the auxiliary population Pop2;]]> (3) <![CDATA[Convert the numerical forms of Pop1 and Pop2;]]> (4) forger=1:G%ger is the current iteration number (5) <![CDATA[Auxiliary population Pop2 update: Random strategy update;]]> (6) <![CDATA[Update of main population Pop1: Select learning objects from Pop2 and guide the particles to update their positions;]]> (7) Optimize frontier distribution; (8) <![CDATA[If the interaction condition between Pop2 and Pop1 is satisfied, exchange some individuals between Pop2 and Pop1;]]> (9) <![CDATA[Arrange Pop1 and Pop1;]]> (10) <![CDATA[Perform numerical form conversion and non-dominated sorting on Pop1;]]> (11) end (12) Output the optimal solution

[0144] 2.1 Encoding and Decoding

[0145] The embodiment of the present invention aims to optimize the three objective variables of task allocation, task scheduling sequence after allocation, and replenishment time. The PSO algorithm is used to solve the problem, and a particle represents a solution to the problem. In order to simplify the decoding process, integer encoding is used, and the particle structure is a two-stage representation method. The first stage is a nested array, whose size is a 1*I matrix, indicating that there are I inner arrays, and the i-th column array X 1i The inner size is 1*n i , represents the rescue mission assigned to UAV i and its execution order, n i is the number of rescue missions assigned to UAV i; suppose there are J rescue missions in total, so X 1i The value range is [1,J]; the second stage is 1*ns i The array X 2i , represents the resupply time of UAV i, which is represented by the rescue mission number, that is, the UAV will be resupplied after completing the mission, ns i is the number of times UAV i needs to be resupplied during its mission, and its value range is {1,2,…,n i -1};

[0146] In the decoding stage: a two-stage embedding method is used, that is, in the first stage, the starting point, the replenishment point, and the end point are embedded according to the array content in the second stage; if the i-th particle X i To decode, first X i The jth column of the first stage is decoded to separate the task allocation of the i-th UAV, that is, a series of task points, represented by X i ; According to the supply point location information given at the corresponding position in the second stage, after the supply point is embedded in the task, the embedded element is a larger integer M (the element embedded in the present invention is the supply point sequence number plus 1000) to distinguish it from the rescue task; the embedding operation follows the proximity principle, that is, the supply point closest to the first rescue task to be performed after the drone is replenished is selected, and finally the base (represented by 0) is embedded at both ends of the drone task allocation sequence, thereby obtaining the decoded drone's real flight path, which includes the starting point, rescue point, supply point and end point.

[0147] Take the particles shown in Table 2 as an example, which represents a scenario where 3 drones perform 12 rescue missions. Assume that there are two resupply points in this scenario, represented as 1001 and 1002. If the rescue mission assigned to drone 1 is [2, 5, 9, 6], it needs to be resupplied after completing the second rescue mission. The resupply point closest to mission 5 is 1002, which is selected as the resupply point. Therefore, its path is O1 = [0, 2, 5, 1002, 9, 6, 0]; similarly, O2 = [0, 8, 4, 1001, 5, 7, 1002, 12, 0], O3 = [0, 11, 10, 0].

[0148] Table 2 Example of two-stage particle encoding

[0149] Drone Number <![CDATA[UAV1]]> <![CDATA[UAV2]]> <![CDATA[UAV3]]> Assigned tasks (Phase 1) [2,5,9,6] [8,4,1,5,7,12] [11,10] Resupply Time (Phase 2) [2] [2,5] [\]

[0150] In addition, regarding the specific amount of supplies that the drone needs to replenish when it is at the supply station, the maximum value required to complete the rescue mission is automatically set; for example, O1 = [0, 2, 5, 1002, 9, 6, 0], the maximum demand quantities obtained by rescue point 2 and rescue point 5 based on the number of trapped people are 50 and 10 respectively, and the amount of supplies carried by drone 1 when it departs from the base is 60; the material demand quantities obtained by rescue point 9 and rescue point 6 based on the number of trapped people are 30 and 20 respectively, so the amount of supplies that drone 1 needs to replenish at supply station 1002 is 50.

[0151] 2.2. Problem feature-guided initialization strategy

[0152] As mentioned above, the variables that need to be optimized in this invention are mainly: the task allocation scheme of the drone, the task sorting, and the selection of the resupply time, which correspond to the first and second rows of particles respectively. Therefore, the proposed algorithm is mainly divided into three steps when generating the initialization scheme: (1) assigning all tasks to drones; (2) sorting the tasks assigned to each drone; (3) selecting the time when resupply is required during the execution of the task. In view of the above steps, this section gives specific strategies.

[0153] 2.2.1. Clustering-based initialization task allocation

[0154] For the UAV task allocation problem, the flight distance is an important evaluation index. Therefore, when the UAV continuously performs multiple tasks with similar distances, the flight distance can be shortened to obtain better results. To this end, the present invention proposes an initialization task allocation strategy based on K-means, which is used to select the value of the first row of particles.

[0155] First, based on the location information of N rescue missions, the K-means clustering algorithm is used to divide these N tasks into I subclusters; then, the I subclusters are randomly assigned to I drones, thereby completing the initial task assignment. In particular, in order to prevent the same clustering result from being assigned multiple times, resulting in serious particle homogeneity, the present invention sets an external archive set Ar1 to save different clustering results. In addition, the scale of Ar1 is limited by setting the number of clustering algorithm runs, and the value is generally taken as the PSO population size. When initializing the allocation, the clustering results in Ar1 are called cyclically, and each cycle will assign the clustering schemes of different results in Ar1 to the particles in the population that have not yet been assigned tasks, one by one, to avoid particle homogeneity to the greatest extent. The schematic diagram of particle initialization is shown in the figure. Figure 4 shown.

[0156] 2.2.2. Task sorting based on penalty or time trend

[0157] After completing the initial task assignment in Section 2.2.1, the next step is to sort the rescue tasks of each drone in the particle. This section presents an evaluation metric that takes into account both rescue value (f1) and distance (f2) to calculate the comprehensive benefits of different tasks. Assuming that rescue task j is one of the rescue tasks that drone i is responsible for, the comprehensive benefit of selecting rescue task j as the next rescue task performed by drone i is:

[0158]

[0159] in, represents the comprehensive benefit of drone i selecting rescue mission j as the next rescue mission at time t; v j (t) represents the rescue value of rescue mission j; represents the time it takes for drone i to arrive at rescue mission j from the base; d i,j represents the distance between rescue mission j and drone i; α represents the weight.

[0160] 2.2.3. Probabilistic replenishment strategy based on load

[0161] After completing the task allocation and sorting operations, it is necessary to choose the right time to resupply the drone under the current task execution order. On the one hand, it can prevent the drone from being unable to complete the assigned task due to cargo capacity restrictions. On the other hand, by increasing a certain distance in exchange for increased speed, it can reach the rescue point faster.

[0162] According to the relationship between the drone load and speed in formula (5), the greater the load, the slower the drone’s flight speed. When the drone is empty, its flight speed is the maximum, denoted as vel max ; When fully loaded, the UAV’s flight speed is the minimum, recorded as (1-ε)×vel max , where ε is the influence coefficient, which is determined by the performance of the drone itself. Assume that after drone i completes the jth assigned task, the total amount of materials it has unloaded is At this time, the supply probability of UAV i can be expressed as shown in formula (16):

[0163]

[0164] in, represents the probability of resupply of UAV i after completing the jth rescue mission; S max represents the maximum cargo capacity of the drone; in particular, when and , it means that UAV i cannot complete the j-th rescue mission assigned to it. In this case, after completing the j-1-th rescue mission, the UAV must be resupplied.

[0165] 2.3 Main population evolution

[0166] After the algorithm is initialized, the initial population Pop is sorted in the form of interval numbers, thereby obtaining a new population Pop' with new particle sorting. Then, the new population Pop' is divided into the main population Pop1 and the auxiliary population Pop2 according to the ratio θ. Specifically: First, according to the principle of the least common multiple of integers, the minimum integer θ that can be divided according to the ratio θ is determined. int , and then select the first θ from the new population Pop' according to the non-dominated sorting result int The particles form a minimum loop unit, and the particles are distributed to Pop1 and Pop2 in the loop unit according to the ratio of θ. After that, the next θ is selected in sequence. int The process is repeated until all particles in Pop' are allocated. Figure 5 In the equation, θ = 2 / 1, θ int =3, according to θ int Loop and distribute the individuals in Pop to Pop1 and Pop2 in proportion.

[0167] After the population is divided, the main population and the auxiliary population adopt a simultaneous update strategy. In the main population Pop1, unlike the traditional PSO update strategy, the setting of the global extreme value is canceled to avoid excessive concentration and homogeneity of the population. The algorithm introduces the auxiliary population Pop2, i.e., the learning library, to select learning objects for speed update. The specific update method adopts the method in the document "Refinement of Collaborative Multitasking and Multi-Objective Allocation in Heterogeneous UAV Swarms An Approach Informed by Decomposition Learning and Particle Swarm Optimization" (hereinafter referred to as Document 2).

[0168] 2.3.1. Selection of learning objects

[0169] The present invention selects learning objects through the following steps: first, determine the particles in the learning library that are not dominated by the particles to be updated as the range of learning objects; then, select the particles with the shortest vertical distance to the line connecting the particles to be updated and the origin within the range as the learning objects. Figure 6 shown.

[0170] Figure 6 In (a), the solid dots represent particles, red represents the objects to be updated in the main population, green represents the range of learning objects in the learning library that meet the dominance relationship, and the number below the particle is its dominance level. By connecting the origin and the straight line L of the particle to be updated, the vertical distance between each green particle and L is calculated, and the particle with the shortest distance is selected as the learning object. Therefore, Figure 6 In (a), the green particle marked with a yellow star is the learning object of the red particle. This method can prevent the population from being overly concentrated on a single particle, thereby preventing the algorithm from falling into a local optimal solution. Figure 6 (b) shows an example of using this method to select learning objects. It can be seen that not all particles on the non-frontier select particles on the frontier as learning objects, thus ensuring the diversity of the population in iterative updates.

[0171] In addition, the above method is used to select learning objects, which helps high-quality particles have a greater advantage when they are selected as learning objects, thereby accelerating the convergence of the algorithm. This can be illustrated by the following proof:

[0172] First, define the similarity: the absolute value of the difference in f1 / f2 between two particles A and B is the similarity, as shown in formula (17):

[0173]

[0174] Assume that X1 is to be updated, X2 and X3 are candidate learning objects, and X2 dominates X3. Figure 7 , assuming that X2 and X3 have the same similarity with X1 in the target space, that is, ∠X2Of1=∠X3Of1, then △X2OA∽△X3OB. According to the properties of similar triangles, |OX2|<|OX3|, and thus |AX2|<|BX3|. According to the selection criteria for learning objects, X1 is more likely to select particle X2 as a learning object. In addition, since |AX2|=|BC|, when X3 is located in the triangular area CX2X3, that is, X3 is more similar to X1 than X2, but |BX3|>|AX2| always holds, so X2 has a higher priority. This class selection mechanism makes it easier for high-quality particles to be selected as learning objects, thereby accelerating the convergence of the algorithm while ensuring diversity.

[0175] 2.3.2 Particle learning update steps

[0176] After completing the selection of the learning object, the update process of the particles to be updated will be divided into two stages, using the method proposed in Reference 2 for iterative updating and correction. Figure 8The example illustrates the entire update process: After the particle completes the learning operation, the component UAV2 marked with a green background represents the new allocation scheme. However, this scheme conflicts with UAV1 and UAV3 (marked with a red background), resulting in repeated execution of tasks 6 and 8, and unallocated tasks 1 and 7 (marked with a blue background), which puts the particle in an infeasible position. At the same time, the infeasible particle is corrected (the involved drones except the learning drone (UAV2) are marked with a yellow background), so that the particle is transferred from the infeasible area to the feasible area, and the update is completed.

[0177] 2.4. Learning library evolution based on auxiliary population

[0178] This section introduces the update method of the auxiliary population Pop2 and its auxiliary role in the main population Pop1. As a learning library, the auxiliary population Pop2 updates itself through learning, mutation and screening, and interacts with the main population Pop1.

[0179] 2.4.1. Self-update method within the learning library

[0180] The learning strategy in Section 2.3 is used as the update operator Δ1 for updating individuals in the learning library, and the genetic mutation operator Δ2 proposed in Reference 2 is introduced. When updating, the auxiliary population Pop2 is divided into two blocks, and each block is applied with a different update operator. The range of the block is determined randomly, and the specific division is based on the random number rank scope The random number calculation method for determining the non-dominated level is shown in formula (18).

[0181] rank scope =[1,rand([1,rank max / 2])] (18)

[0182] Among them, rank scope Indicates the non-dominated rank range of the selected population; rank max is the maximum non-dominated rank of the current era; rand([1,rank max / 2]) is [1,rank max / 2] between random numbers. In the selected range (non-dominated level is less than rank scope ) uses the genetic mutation operator Δ2 to update the position, which helps the algorithm to jump out of the local optimum and search for more solutions. scope ) is updated through the update operator Δ1, which helps to accelerate the convergence of the population.

[0183] When using the genetic mutation operator Δ2 to mutate, the guiding direction needs to be selected according to the individual position. The present invention divides the individual space into two parts and sorts the individuals according to the f1 / f2 value. The individuals in the first half are mutated with the optimization of f1 as the guide, and the individuals in the second half are mutated with the optimization of f2 as the guide. During the mutation process, the task sorting strategy based on the penalty (f1) or time (f2) trend mentioned in Section 2.2.2 is adopted to achieve the goal by adjusting the weight parameter α. Fig. 9 The evolution process is shown: the blue and black individuals are the individuals to be updated, where the blue individuals are updated using the Δ2 operator determined by formula (18), and the blue arrow indicates the direction of their mutation. The green individuals are high-quality individuals that can replace the original individuals after mutation, and the red individuals are low-quality individuals that cannot replace the original individuals. The light blue area indicates the area guided by the optimization of f1, and the light yellow area indicates the area guided by the optimization of f2. The black individuals are updated using the update operator Δ1, and the black arrow indicates the learning object of the individuals to be updated. This method enhances the extension of the Pareto frontier while ensuring convergence.

[0184] 2.4.2 Interaction between the learning library and the main population

[0185] Since the size of the learning library is limited and new individuals are intermittently absorbed from the main population during the algorithm iteration update process, it is necessary to introduce an elimination machine to maintain the size of the learning library unchanged.

[0186] According to Section 2.3, when individuals in the main population select learning objects, they will determine the learning objects based on the non-dominated sorting results of the population within the learning library and the position of the individual itself. Therefore, non-dominated sorting must have been performed within the learning library, and the elimination mechanism is also based on the non-dominated sorting results.

[0187] In each iterative update of the learning library, new individuals are generated by the update operator Δ1 or the genetic mutation operator Δ2. The new individuals generated by the update operator Δ1 will determine whether to replace the old individuals based on the fitness value, while the new individuals generated by the genetic mutation operator Δ2 will participate in the non-dominated sorting and crowding sorting together with the old individuals, and eliminate the old individuals with the lower sorting, and the number of eliminated individuals is equal to the number of newly added individuals, so as to keep the size of the learning library unchanged.

[0188] In addition, in order to maximize the use of the learning library, the exchange frequency between the population inside the learning library and the main population is controlled by setting the parameter τ. By default, the value of τ is 1 / 10 of the maximum number of iterations G of the algorithm.

[0189] 3. Simulation experiment

[0190] 3.1 Evaluation Indicators

[0191] The model proposed in this paper is an uncertain model containing data in the form of interval numbers, where f1 in the target space is an interval number and f2 is an exact number. In the algorithm iteration process, the numerical processing method proposed in Section 1.4 is adopted to perform non-dominated sorting through a mixed form of interval numbers and exact numbers. The resulting solution is represented in the target space as follows Fig.10 As the traditional HV value calculation method cannot accurately reflect the quality of the solution, the present invention uses double HV value (Double Hypervolume, DHV) to evaluate the quality of the solution. Fig.10 The solution in (a) calculates the DHV value as Fig.10 (b) as shown.

[0192] exist Fig.10 In (a), the red dot is the reference point [f1(r*), f2(r*)]; the particle Xu is composed of interval numbers and exact numbers, and the coordinates of the corresponding target space are ([f1(Xu1 - ),f1(Xu1 + )],f2(Xu2)), (change the position of the superscripts and subscripts) Therefore, due to the uncertainty of f1, the HV value cannot be accurately obtained. The HV value range of particle Xu can only be determined in region B, but not in regions A and C.

[0193] Therefore, the present invention adopts Fig.10 (b) is used to calculate the DHV value. In the figure, the red points are reference points, and the solutions are divided into two categories: the lower limit solution composed of blue points and the upper limit solution composed of yellow points. The HV values ​​of the upper and lower limit solutions are calculated and added, as shown in formula (19)

[0194]

[0195] The area of ​​the blue area represents the HV value of the lower limit solution, the area of ​​the yellow area represents the HV value of the upper limit solution (the yellow area in the figure is completely covered), and the green area is the overlap of the two. Using the DHV value to evaluate the quality of the solution can take into account the influence of the location and uncertainty of the solution on its quality, which is suitable for the algorithm proposed in this invention.

[0196] 3.2 Experiment and parameter setting

[0197] In order to verify the superiority of the PILPSO algorithm proposed in this invention, a comparative experiment was conducted. The specific scenario settings are shown in Table 3. The effectiveness of the key strategies was verified based on 30 experimental results. The comparative experiment selected four advanced multi-UAV task allocation algorithms (AAUC-IBEA, CCPSO, MODABC and SLPSO) for comparison under the same scenario. In the experiment, the DHV value mentioned in Section 3.1 was used as the performance evaluation index. The larger the DHV value, the better the algorithm performance. In addition, the Wilcoxon rank-sum test was used to determine whether the PILPSO algorithm was significantly better than the comparison algorithm, and the confidence level was set to 0.05. Among them, "--" and "++" indicate that the PILPSO algorithm is significantly inferior to and better than the comparison algorithm, respectively, and "≈" indicates that there is no significant difference between the two.

[0198] Table 3 Scene settings

[0199]

[0200] 3. Experimental results and analysis

[0201] In this experiment, the proposed PILPSO algorithm is compared with four comparison algorithms: AAUC-IBEA, CCPSO, MODABC[ and SLPSO, respectively, in three scenarios. Table 4 summarizes the results of the five algorithms in three scenarios.

[0202] Table 4 Comparative experimental DHV values

[0203]

[0204] From the results of scenario 1, it can be seen that the performance of the PILPSO algorithm proposed in the present invention is not superior to that of MODABC and SLPSO in simple scenarios; however, with the increase of scenario complexity, such as in scenarios 2 and 3, the performance advantage of the PILPSO algorithm gradually emerges, and in these two scenarios it is completely superior to the other four comparison algorithms, proving that the PILPSO algorithm has a strong optimization ability.

[0205] For further analysis, FIG11 shows a box plot of DHV values ​​of the method proposed in the present invention and four algorithms in three scenarios.

[0206] As shown in Figure 11(a), in scenario 1, the PILPSO algorithm performs better, second only to the MODABC algorithm.

[0207] As shown in Figure 11(b), in scenario 2, PILPSO performs best, with a clear advantage over the comparison methods.

[0208] As shown in Figure 11(c), in scenario 3, the scenario is more complex and the advantage of the PILPSO algorithm is more obvious.

[0209] Figure 11 again verifies that as the scene complexity increases, the PILPSO algorithm shows a more obvious advantage in performance compared with the other four comparison algorithms. The data in Table 4 and Figure 11 both prove that the algorithm proposed in the present invention performs better in complex scenes.

[0210] In order to further study the performance of the algorithms, three scenarios were selected to analyze the Pareto front distribution of the five algorithms in the experiment closest to the mean among the 30 experiments.

[0211] As shown in Figure 12(a), in scenario 1, due to the low complexity of the scenario and the small feasible domain, the Pareto front distribution overlap of the five algorithms is high. PILPSO performs well in distribution and convergence, while MODABC is better than PILPSO in convergence but slightly worse in distribution. The other three comparison algorithms are inferior to PILPSO in convergence and distribution.

[0212] As shown in Figure 12(b), in scenario 2, as the complexity of the scenario increases, the performance advantage of PILPSO gradually emerges. It is completely superior to the other four comparison algorithms in terms of convergence. In terms of distribution, except for MODABC, PILPSO is superior to the other three comparison algorithms.

[0213] As shown in Figure 12(c), in scene 3, which has the highest scene complexity, the performance gap between PILPSO and the other four comparison algorithms is further widened. In terms of convergence, PILPSO shows an absolute advantage; in terms of distribution, although MODABC has a slight advantage in distribution range, the other algorithms are inferior to PILPSO in terms of distribution range and distribution uniformity. At the same time, by observing the width of the particles on the Pareto front of the five algorithms (indicating the uncertainty of the solution), it can be seen that the solution uncertainty of PILPSO and MODABC is small, while the uncertainty of the other three comparison algorithms is large.

[0214] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0215] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined in the present invention may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown in the present invention, but will conform to the widest scope consistent with the principles and novel features disclosed in the present invention.

Claims

1. A method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm, characterized in that: The steps include: Based on the rescue value of each rescue mission and the time it takes for each drone to arrive at the rescue mission it is responsible for from the base, determine the minimum rescue comprehensive penalty objective function; Determine the shortest rescue time objective function based on the time it takes each drone to reach the last rescue mission it is responsible for; According to the minimum rescue comprehensive penalty objective function and the shortest rescue time objective function, a UAV cluster collaborative rescue task allocation model is constructed; The UAV cluster collaborative rescue task allocation model is solved based on the population interactive learning particle swarm algorithm, and the UAV cluster collaborative rescue tasks are allocated based on the solution results.

2. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 1 is characterized in that: The minimum rescue comprehensive penalty objective function is expressed as: Among them, f1 represents the minimum rescue comprehensive penalty objective function; I represents a total of I drones; i represents the i-th drone; J represents a total of J rescue missions; j represents the j-th rescue mission; represents the time it takes for drone i to arrive at rescue mission j from the base; Represents the rescue value of rescue mission j at time t.

3. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 2 is characterized in that: The time when each drone arrives at the rescue mission from the base, and the corresponding acquisition steps include: Based on the maximum empty-load flight speed of the drone, the maximum cargo capacity of the drone, and the amount of rescue materials each drone carries before arriving at each rescue mission, the flight speed of each drone between every two rescue missions is obtained; it is expressed as: Among them, vel i,j represents the flight speed of UAV i between rescue mission j-1 and rescue mission j; rescue mission j-1 is the previous rescue mission of rescue mission j; ε represents the UAV performance impact coefficient; vel max Indicates the maximum flight speed of the drone without load; S max Indicates the maximum cargo capacity of the drone; S i,j represents the amount of rescue supplies of UAV i before it arrives at rescue mission j; Based on the flight speed and distance of each drone between each rescue mission, the time it takes for each drone to arrive at the rescue mission it is responsible for from the base is obtained, expressed as: in, J represents the time it takes for UAV i to arrive at rescue mission j from the base; i represents the number of rescue missions performed by UAV i from its departure from the base to its return to the base, and L(j,j-1) represents the distance between rescue mission j-1 and rescue mission j; ΔT represents the sum of the unloading time and replenishment time of drone i.

4. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 1 is characterized in that: The shortest rescue time objective function is expressed as: Among them, f2 represents the shortest rescue time objective function; represents the time when UAV i arrives at the last rescue mission it is responsible for; I represents the total number of I UAVs; ΔT represents the sum of the unloading time and the replenishment time of UAV i.

5. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 1 is characterized in that: The encoding and decoding strategies of the population interactive learning particle swarm algorithm specifically include: In the encoding stage: integer encoding is used, and the particle structure is represented by a two-stage method; the first stage is a nested array, whose size is a 1*I matrix, indicating that there are I inner arrays, and the i-th column array X 1i The inner size is 1*n i , represents the rescue mission assigned to UAV i and its execution order, n i is the number of rescue missions assigned to UAV i; suppose there are J rescue missions in total, so X 1i The value range is [1,J]; the second stage is 1*ns i The array X 2i , represents the resupply time of UAV i, which is represented by the rescue mission number, that is, the UAV will be resupplied after completing the mission, ns i is the number of times UAV i needs to be resupplied during its mission, and its value range is {1,2,…,n i -1}; In the decoding stage: a two-stage embedding method is used, that is, in the first stage, the starting point, the replenishment point, and the end point are embedded according to the array content in the second stage; if the i-th particle X i To decode, first X i The jth column of the first stage is decoded to separate the task allocation of the i-th UAV, that is, a series of task points, represented by X i ; According to the supply point location information given by the corresponding position in the second stage, the supply point is embedded in the task; the embedding operation follows the principle of proximity, that is, the supply point closest to the first rescue task to be performed after the drone is resupplied is selected, and finally the base is embedded at both ends of the drone task allocation sequence, thereby obtaining the decoded drone's real flight path, which includes the starting point, rescue point, supply point and end point.

6. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 1 is characterized in that: The particle swarm algorithm based on population interactive learning is used to solve the drone cluster collaborative rescue task allocation model, specifically including: Initialize the population interactive learning particle swarm algorithm based on K-means clustering to obtain the initial population; Divide the initial population into a main population and an auxiliary population; Determine the particles to be updated in the main population, and select particle learning objects based on the auxiliary population; Based on the particle learning object, the particles to be updated are updated.

7. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 6 is characterized in that: The K-means clustering-based population interactive learning particle swarm algorithm is initialized to obtain an initial population, specifically including: (1) Based on the location information of J rescue tasks, the K-means clustering algorithm is used to divide the J rescue tasks into I sub-clusters; I sub-clusters are randomly assigned to I drones, thereby completing the initial task assignment; (2) Based on the comprehensive benefits corresponding to the rescue missions that each drone is responsible for, the rescue missions of each drone are ranked; specifically: let rescue mission j be one of the rescue missions that drone i is responsible for, and the comprehensive benefit of selecting rescue mission j as the next rescue mission to be performed by drone i is: in, represents the comprehensive benefit of drone i choosing rescue mission j as the next rescue mission at time t; v j (t) represents the rescue value of rescue mission j; represents the time it takes for drone i to arrive at rescue mission j from the base; d i,j represents the distance between rescue mission j and drone i; α represents the weight; (3) Under the current rescue mission execution order of UAV i, resupply the UAV; specifically: suppose that after UAV i completes the assigned j-th rescue mission, the total amount of rescue supplies unloaded by UAV i is At this time, the supply probability of drone i is expressed as: in, represents the probability of resupply of UAV i after completing the jth rescue mission; S max represents the maximum cargo capacity of the drone; in particular, when and , it means that UAV i cannot complete the j-th rescue mission assigned to it. In this case, after completing the j-1-th rescue mission, the UAV must be resupplied.

8. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 6 is characterized in that: The dividing of the initial population into a main population and an auxiliary population specifically includes: Perform non-dominated sorting in the form of interval numbers on the initial population Pop to obtain a new population Pop' with new particle sorting; According to the principle of the least common multiple of integers, determine the smallest integer θ divided by the ratio θ int ; According to the non-dominated sorting result, select the first θ from the new population Pop' int Particles constitute a minimum cycle unit, and in the minimum cycle unit, particles are distributed to the main population Pop1 and the auxiliary population Pop2 according to the proportion θ; continue to select the next θ according to the non-dominated sorting result int The particles are divided and the process is repeated until all particles in the new population Pop' are allocated.

9. The method for allocating UAV cluster collaborative rescue tasks based on population interactive learning particle swarm algorithm according to claim 6 is characterized in that: The determining of the particle learning object based on the main population and the auxiliary population specifically includes: The particles in the auxiliary population that are not dominated by the particles to be updated are taken as the range of particle learning objects; The particle with the shortest vertical distance to the line connecting the particle to be updated and the origin is selected as the particle learning object within the range.

Images