A Multi-UAV Energy Assurance Task Allocation Method

Abstract

The present invention discloses a method for multi-UAV energy guarantee task allocation, including: multi-UAV task allocation service analysis; establishing an objective function for the time required for the UAVs to complete all the assigned tasks; establishing an auxiliary objective function, and respectively using the genetic algorithm to perform iterative optimization on the objective function and the auxiliary objective function; in the process of evolution, using the crossover and mutation operators, generating offspring through selective mating, then evaluating the offspring population through selective imitation, after merging the parent population and the offspring population, sorting according to the factor cost of the merged population, re-determining the factor level, updating the scalar fitness and skill factors of each individual in the merged population, and finally performing environmental selection on the individuals in the population through the elitist strategy to enter the next iteration until the evolution stop condition is met; outputting the multi-UAV energy guarantee task allocation result. The present invention establishes a task allocation model and obtains an effective task allocation scheme through efficient algorithm solving.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicles, and particularly relates to a method for task allocation of energy supply for multiple unmanned aerial vehicles. Background Art

[0002] A high-performance individual combat system urgently demands highly reliable energy supply. In recent years, in order to solve the problem of electric energy supply in battlefield environments, new electric energy supply equipment such as university photovoltaic power generation, small wind power generation, fuel cells, and battery energy storage have been successively equipped for the troops, providing new means for the troops to collect energy locally, store energy locally, and convert energy locally in combat environments. The individual power supply system includes various forms of energy collection means, such as thin-film solar energy, foldable small wind turbines, biomechanical or bionic energy collectors, etc.

[0003] At the same time, significant progress has been made in battery technology. Currently, several individual power supply products have been introduced at home and abroad. Although the energy efficiency of batteries has been continuously improved with the development and progress of battery technology, it still cannot meet the actual power demand during combat. From the development trend of individual combat equipment, major military powers have vigorously developed combat energy supply, mainly focusing on developing new energy sources, researching new power generation technologies, and portable batteries. However, with the emergence of new battlefield characteristics, individual energy supply faces major challenges, mainly in the following aspects: single supply means; few types of individual energy sources; poor reliability of the energy in use.

[0004] Energy occupies a leading position in military revolutionary changes. The improvement of individual energy supply means will bring huge returns to individual confrontation on future battlefields.

[0005] Based on this application background, this chapter conducts research on using multiple unmanned aerial vehicles to deliver batteries to provide battlefield individual energy supply. The primary task of wartime military energy supply is to meet the energy demand of combat operations. Under conditions such as harsh environments, low safety factors, and inconvenient ground transportation, the miniaturization and intelligence of energy support and supply equipment are an inevitable trend in the development of the future energy supply equipment system. Using unmanned aerial vehicles for military energy supply is not only an enrichment of supply means but also greatly improves the supply efficiency.

[0006] At the same time, as a brand-new combat method, the cooperative combat of unmanned aerial vehicle clusters will surely play an important role in future joint operations.

[0007] Generally, a formation of unmanned aerial vehicles consists of multiple types of unmanned aerial vehicles, which carry different payloads and weapon equipment systems, and their platform performances also vary. Therefore, each unmanned aerial vehicle has different task execution capabilities when performing different types of tasks.

[0008] At present, the most typical small unmanned aircraft in China is the multi-rotor unmanned aircraft, with a flight speed of up to 50 km / h and an effective payload of up to 20 kg. It is very suitable for the emergency support of small materials such as blood and ammunition urgently needed during wartime. Therefore, using this type of unmanned aircraft to deliver batteries in individual soldier energy support to meet the energy needs of the individual soldier combat system has significant advantages, which are mainly manifested in the following aspects:

[0009] (1) Unmanned aircraft delivery can quickly respond to battlefield needs and achieve precise and efficient support

[0010] (2) Low delivery cost, strong emergency response ability, and can achieve uninterrupted operation

[0011] (3) Few constraints, suitable for high-difficulty tasks, and can achieve zero casualties

[0012] (4) Enhance the anti-destruction and robustness of end delivery

[0013] However, at present, the military's research theories and applications on unmanned aircraft delivery are relatively few, and carrying out unmanned aircraft delivery under special conditions has certain significance for military support during wartime. Summary of the Invention

[0014] In view of this, the present invention proposes a multi-unmanned aircraft energy support task allocation method, using unmanned aircraft to deliver batteries to solve the challenges faced by individual soldier energy support on the current battlefield, designing the architecture of the battlefield unmanned aircraft delivery system, starting from the perspective of the operation process, establishing an unmanned aircraft task allocation model and solving it with an algorithm to quickly respond to the energy needs of individual soldiers and provide a support plan.

[0015] The multi-unmanned aircraft energy support task allocation method disclosed by the present invention includes the following steps:

[0016] Multi-unmanned aircraft task allocation service analysis, using a triple {I, T, C} to describe the multi-unmanned aircraft task allocation scenario, where I = {I1, I2,..., I m} is the set of unmanned aircraft, indicating that there are m unmanned aircrafts executing tasks on the battlefield, T = {T1, T2,...T n} is the task set, indicating that there are n tasks to be executed on the battlefield, and C represents the constraint conditions in the scenario;

[0017] Establish the objective function of the time required for the unmanned aircraft I i to complete all the assigned tasks;

[0018] Simulate and sample the objective function through the Monte Carlo method to obtain sample points, and then establish an auxiliary objective function by fitting the sample points with a Gaussian function. Use the genetic algorithm to iteratively optimize the objective function and the auxiliary objective function respectively;

[0019] Initialize the population according to the set algorithm parameters, and calculate the factor cost of this population under each iterative optimization task, that is, the objective function value;

[0020] After sorting the population list in ascending order of factor cost, mark the sequence index of the individuals in the population list, that is, the factor level. When multiple individuals have the same factor cost, use the random tie-breaking method;

[0021] Calculate the scalar fitness of each individual according to the factor level and determine the skill factor of this individual;

[0022] During the evolution process, use the crossover and mutation operators, generate offspring through selective mating, then evaluate the offspring population through selective imitation. After merging the parent population and the offspring population, sort according to the factor cost of the merged population, re-determine the factor level, update the scalar fitness and skill factors of each individual in the merged population, and finally perform environmental selection on the individuals in the population through the elite strategy to enter the next iteration until the evolution stop condition is met;

[0023] Output the task allocation result of multi-UAV energy guarantee.

[0024] Further, the UAV I i The time required to complete all the assigned tasks is as follows:

[0025] Use two two-dimensional matrices A and B to represent the task allocation scheme. Among them, matrix A is a natural number matrix, and the element a in matrix A ij represents the order of UAV I i to execute task T j The element b in matrix B is an integer matrix, and the element b in matrix B ij represents the number of batteries that UAV I i supports and guarantees to the task point T j ;

[0026] Calculate the total number of tasks that each UAV needs to execute from matrix A, denoted as count i ; Use to represent the task sequence that UAV I i needs to execute, where the value of each element d ik represents the kth task that UAV I i needs to execute is task F i represents the time required for UAV I i to complete all the assigned tasks, including the travel time and the delivery time. The calculation formula of F i is as follows:

[0027]

[0028] Among them, represents the distance from the warehouse W to the mission i for the drone I distance. represents the distance from the mission i for the drone I to the mission distance. represents the distance from the warehouse W to the mission i for the drone I distance. is the delivery time for the drone I i V i is the flight speed of the drone I i flight speed.

[0029] Furthermore, there are the following constraint conditions:

[0030] The drone I i load constraint during the mission execution, that is:

[0031]

[0032] is the maximum load of the drone I i maximum load;

[0033] The drone satisfies the range limit constraint during the mission execution, that is:

[0034]

[0035] S i is the farthest range of the drone I i farthest range;

[0036] The power supply request of each demand point must be satisfied, so:

[0037]

[0038] Q j is the number of battery cells required at the mission point.

[0039] Furthermore, the objective function of the following objective optimization model is established:

[0040] minimize maxmize(F i )

[0041]

[0042] In the current mission assignment scenario, the objective of the model is to minimize the project duration for task completion, that is, the time taken by the last drone to complete its mission sequence, that is, the makespan. The objective function is:

[0043]

[0044] Furthermore, the Gaussian function is defined as follows:

[0045]

[0046] where d is the number of decision variables, a and b are parameters to be estimated, x i represents the i-th variable, μ i and σ i are obtained by minimizing the error between the estimated value of the Gaussian function and the actual value of the sample;

[0047] To fit the function y = f(x), x ∈ R d , n sampling points x = (x 1 , x 2 ,..., x n ) ∈ R d and their corresponding observed values y = (y 1 , y 2 ,..., y n ) are sampled from the original function. For any two sampling points x i , x i ′ ∈ R d , the correlation c between them is defined as:

[0048]

[0049] where 1 ≤ p i ≤ 2 is used to measure the smoothness of the fitting function, θ i ≥ 0 represents the importance of x i to the fitting function f(x). Then, the hyperparameter values of the Gaussian function are obtained by minimizing the error function, and the error function is defined as follows:

[0050]

[0051] where C is an n×n matrix composed of c(x, x′), and I is an n×1 unit vector.

[0052] Furthermore, in the way of integer coding, the task serial number is used as the chromosome gene;

[0053] Each time a single soldier requests energy replenishment, information about a demand for 30 batteries will be sent to the control center to sustain the electrical energy required for the next two days of operations. A demand of 60 at a certain mission point means that two single soldiers at that location have sent energy replenishment requests. This task is decomposed into two subtasks with a demand of 30 each. By this method, all tasks in the task list are disassembled into multiple subtasks with a demand of 30, and each mission point can only be visited once.

[0054] Furthermore, false tasks are set up for supplementation to ensure that each drone is assigned to multiple mission points and to facilitate subsequent genetic operations; the false tasks are numbered with negative integers and serve as chromosome genes.

[0055] Furthermore, the mutation process of the genetic algorithm is as follows: First, randomly select a certain position as the mutation point a, and the gene at this mutation point is denoted as g1; second, randomly generate a gene encoding within a certain range, denoted as g2; then find the position of gene g2 on the chromosome as the mutation point b; finally, replace the gene at mutation point a with g2 and replace the gene at mutation point b with g1.

[0056] Furthermore, the extraction technique is used to simplify the objective function, that is, randomly select k components, and each component is numbered as [d1, d2,...,[d k .

[0057] The beneficial effects of the present invention are as follows:

[0058] The present invention conducts research on multi-drone task allocation, selects the battlefield single-soldier energy support scenario, establishes a task allocation model based on the actual application requirements, and obtains an effective task allocation plan through efficient algorithm solving. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Flowchart of the task allocation method of the present invention;

[0060] Figure 2 Task distribution map;

[0061] Figure 3 Example diagram of the chromosome gene encoding of an individual;

[0062] Figure 4 Task allocation plan diagram;

[0063] Figure 5 Convergence curve of the present invention in multi-drone task allocation. DETAILED DESCRIPTION OF THE INVENTION

[0064] The present invention will be further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any transformation or replacement based on the teachings of the present invention falls within the protection scope of the present invention.

[0065] The UAV delivery system architecture in the present invention is built in a hierarchical mode, which is also called a multi-layer architecture mode. The hierarchical method satisfies the idea of "high cohesion and low coupling".

[0066] Based on this system design idea, the UAV delivery system is decomposed into three layers: the data layer, the business layer, and the presentation layer. The data layer includes data collection and storage. It mainly obtains battlefield environment, individual soldier positioning, and demand information through sensors, GPS, individual soldier communication modules, etc., and preprocesses and stores the collected data in the database. The business layer is the UAV task assignment system, which contains a task assignment model, mainly including a sub-module - the task assignment module. The core solution algorithm is the multi-factor evolutionary algorithm based on auxiliary objectives. The task assignment module generates a task assignment plan according to the requirements obtained from the data layer. The presentation layer is the interaction interface, which is operated by decision-makers with management authority and can view relevant information and issue scheduling instructions.

[0067] Above, the UAV task assignment in the business layer is an important part of the UAV delivery system. It mainly refers to reasonably allocating targets with different positions, values, and threat levels to UAVs with different types, values, and combat capabilities during the entire process of executing combat tasks, on the premise of meeting tactical and technical indicators, combat task requirements, platform and weapon performance constraints, UAV tactical usage conditions, etc. An efficient task assignment algorithm is used to determine the task execution plan for the UAVs, so as to maximize the overall combat effectiveness and minimize the cost. The present invention studies multi-UAV task assignment and solves it based on the auxiliary target technology.

[0068] Describe the multi-UAV task assignment scenario with a triple {I, T, C}, and the map range is 100km * 100km. Among them, I = {I1, I2,..., I m} is the set of UAVs, indicating that there are m UAVs executing tasks in the battlefield. Each element I i (i = 1, 2,..., m) in the set contains the flight speed V i of the UAV I i , position the maximum range S i , and the UAV type P i . In this scenario, there are 3 types of UAVs, denoted as P = {1, 2, 3}, P i ∈ P, P i = 1 represents that the UAV I i is a type-I UAV, P i = 2 represents that the UAV Ii It is a type-II UAV, P i = 3 represents UAV I i It is a type-III UAV. UAVs of different models have different payloads and startup costs, as detailed in Table 1; T = {T1, T2,... T n} is the task set, indicating that there are n tasks in the battlefield that need to be executed. Each element T j {j = 1, 2,..., n} in the set includes the location of each task and the number of battery cells Q required at the task point j ; C represents the constraints in the battlefield.

[0069] Table 1 Characteristics of heterogeneous UAVs

[0070]

[0071] As the battlefield environment becomes increasingly complex and combat tasks become increasingly diverse, the possibility of UAVs performing tasks alone is getting smaller and smaller. In actual combat, multiple UAVs are generally formed into a formation to cooperate in performing combat tasks. The multi-UAV task allocation problem is a complex constrained optimization problem. The solution space of this problem increases exponentially with the increase in the number of UAVs and tasks. Therefore, this problem is a complex multi-parameter, multi-constraint non-linear continuous optimization problem, and the solution process is complex, requiring a large computational cost and time cost.

[0072] Multi-UAV task allocation service analysis

[0073] A combat unit of an infantry company dispatched a platoon of 30 individual soldiers during a 7-day mission. Each individual soldier carried 30 battery cells when setting out. Under the condition of full-load operation of the individual combat system, 30 battery cells can meet the energy demand for at most 2 days. During the mission, individual soldiers can submit energy supply requests to the command center at any time through satellite equipment. The command center can obtain the real-time location and energy demand information of individual soldiers through the GPS positioning system and communication system, and then provide energy supply by delivering battery cells through UAVs.

[0074] It is known that this combat unit is equipped with 10 type-I UAVs to provide power support. All 10 UAVs are initially parked in the warehouse. In the two-dimensional plane, the coordinates of the warehouse location on the map are defined as the origin (0, 0). Suppose at a certain moment, the command center receives an energy supply request, and the demand information is shown in Table 2. The command center needs to immediately dispatch all UAVs to carry out energy supply according to the request information, and the goal is to complete the energy supplement for individual soldiers in the shortest time.

[0075] Table 2 Task description

[0076]

[0077]

[0078] Based on the above data, the task distribution map at this moment can be drawn, as shown in Figure 3 。

[0079] In addition, in the UAV task allocation scenario described above, the following preconditions also exist:

[0080] A single UAV can only execute one independent task at the same moment;

[0081] The UAV has a constant speed during flight, without considering the acceleration during takeoff and landing, and without considering the influence of load;

[0082] The UAV is in good condition and will not malfunction during the execution of tasks;

[0083] Adopting a centralized control method, there is no communication between UAVs, and they only communicate with the control center;

[0084] The instructions of each UAV are monitored and managed by the decision-making personnel in the control center;

[0085] All UAVs are initially parked in the warehouse. Without considering the gap error during parking, they are all at the coordinate origin;

[0086] The present invention solves the problem of how to allocate tasks to 10 UAVs so that all tasks can be completed in the shortest time.

[0087] Establishment of multi-UAV task allocation model

[0088] The Cooperative Multiple Task Assignment Problem (CMTAP) is mainly used to solve the allocation problem of multi-UAVs collaborating to execute multiple tasks. It is an existing technology in the field and has received attention and continuous optimization and improvement in the field of UAV task allocation. Based on the CMTAP model, the present invention simplifies the problem by task decomposition and introduces resource constraint constraints, restricting the number and flight range of UAVs, with the goal of completing tasks as quickly as possible.

[0089] Setting of the objective function:

[0090] Represents UAV I i To task T j The distance between, Represents task T j To task T j′ The distance between, then UAV I i From warehouse W to task T j The distance is

[0091] The task assignment scheme of the present invention is represented by two two-dimensional matrices A and B. Among them, matrix A is a natural number matrix, and the element a in matrix A ij represents the order in which the drone I_i executes task T j . If a ij is 0, it means that the drone I i does not execute task T j . Matrix B is an integer matrix, and the element b in matrix B ij represents the number of batteries that the drone I i supports and guarantees for the task point T j . Obviously, if a ij = 0, then b ij must be 0.

[0092] The total number of tasks that each drone needs to execute can be calculated from matrix A, denoted as count i . Use to represent the task sequence that the drone I i needs to execute. Among them, the value of each element d ik represents that the kth task that the drone I i needs to execute is task F i represents the time required for the drone I i to complete all the assigned tasks, including travel time and delivery time. The calculation formula of F i is as follows.

[0093]

[0094] Due to the weight limit of the drone payload and the limited weight of the batteries that each drone can carry, during the task execution process, the drone I i needs to meet the load constraint, that is:

[0095]

[0096] At the same time, due to the limited flight distance of the drone itself due to its own endurance problem, during the task execution process, it needs to meet the range limit constraint, that is:

[0097]

[0098] According to the task requirements, the power supply replenishment requests of each demand point must be met. Therefore:

[0099]

[0100] In summary, according to the problem description and target conditions, the present invention establishes the following single-objective optimization model:

[0101] minimize maxmize(F i )

[0102]

[0103] In the current task assignment scenario, the goal of the model is to minimize the project duration for task completion, that is, the time taken by the last drone to complete its task sequence, namely the makespan. The auxiliary goal of the task assignment model in the present invention is designed to minimize the total time for all drones to execute tasks, that is:

[0104]

[0105] Therefore, in the subsequent solution process, the original goal and the auxiliary goal are regarded as two related tasks, and a multi-task optimization framework is used to solve them through a continuous intelligent optimization algorithm.

[0106] Surrogate models commonly used for optimization include Gaussian processes, polynomial regression, radial basis functions, etc. Among them, Gaussian processes are often selected for building surrogate models due to their global nature and smoothness. The reasons for the present invention to select Gaussian processes for the design of the auxiliary goal are as follows:

[0107] (1) Gaussian processes have been proven effective and theoretically reasonable, are simpler than other surrogate models, and have fewer hyperparameters;

[0108] (2) The computational complexity of Gaussian processes is O(N it S 3d ), where N it is the number of iterations, S is the number of sampling points, and d is the number of variables;

[0109] (3) During the optimization process, we only focus on the location of the global optimal solution. Gaussian processes can smooth the shape of the original function and ignore the location of local optimal solutions.

[0110] Mathematically, the Gaussian function is defined as follows:

[0111]

[0112] where d is the number of decision variables, a and b are parameters to be estimated, x i represents the i-th variable, μ i and σ i are obtained by minimizing the error between the estimated value of the Gaussian function and the actual sample value. In order to fit the function y = f(x), x ∈ R d , n sampling points x = (x 1 , x 2 ,..., x n) ∈ R d and its corresponding observed value y = (y 1 , y 2 ,..., y n ). For any two sampling points x, x' ∈ R d , the correlation c between them is defined as:

[0113]

[0114] where 1 ≤ p i ≤ 2, which is used to measure the smoothness of the fitting function, and θ i ≥ 0 represents the importance of x i to the fitting function f(x). Then, the hyperparameter values of the Gaussian function are obtained by minimizing the error function, and the error function is defined as follows:

[0115]

[0116] where C is an n×n matrix composed of c(x, x'), and I is an n×1 unit vector.

[0117] Based on the idea of the surrogate model, the present invention simplifies the original objective function, obtains sample points through Monte Carlo simulation sampling, constructs a function similar to but simpler than the original function through Gaussian fitting, and uses this function as the auxiliary objective.

[0118] Chromosome encoding and decoding method

[0119] In the present invention, integer encoding is adopted, and the task serial number is used as the chromosome gene.

[0120] To simplify the solution model and design an encoding scheme that is easy to evolve, it is necessary to further decompose the tasks in this scenario. Each time a single soldier requests energy replenishment, it will send information about a demand of 30 battery cells to the control center to continue the electrical energy required for the next 2 days of operation. A demand of 60 at a certain task point means that two single soldiers have sent energy replenishment requests at that location. For example, for task T3 in Table 3, we decompose this task into two subtasks with a demand of 30. By this method, all tasks in the task table are disassembled into multiple subtasks with a demand of 30, and each task point can only be visited once.

[0121] Table 3 Task decomposition table

[0122]

[0123] After disassembling the original task, there are 23 tasks in total. Since there are 10 drones and each drone can support at most 3 task points, and can support the energy supply requests of 30 individual soldiers at the same time, the length of the chromosome is 30. And at the beginning, it is assumed that each drone departs fully loaded, so the load constraint can be not considered during the solution process, and finally the actual load carried by the drone is determined according to the number of task points executed.

[0124] It should be noted that false tasks need to be set for supplementation to ensure that each drone is assigned 3 task points and facilitate subsequent genetic operations. The false tasks are numbered with negative integers and used as chromosome genes. The mapping table between gene encoding and tasks is shown in Table 4:

[0125] Table 4 Mapping table between gene encoding and tasks

[0126]

[0127]

[0128] An example of the chromosome gene encoding of an individual is as Figure 3 shown.

[0129] During decoding, the above complete chromosome is evenly divided into 10 segments and disassembled in order. Each segment is a chromosome segment containing 3 genes, representing the task sequence to be executed by a single drone. If there is 1 negative number in this task sequence, it means that the drone only executes 2 tasks except for the negative number; if there are 2 negative numbers, it means that the drone only executes 1 task; if all 3 are negative numbers, it means that the drone does not participate in this energy supply mission.

[0130] Constraint handling method: To meet the maximum flight range constraint of the drone, the ε-punishment method is adopted. If the flight distance of the drone exceeds its maximum flight range, the penalty value ε is added to the fitness of the drone.

[0131] Genetic operator: The OX crossover operator and single-point mutation are used to perform genetic operations to generate offspring.

[0132] The OX crossover operator is as follows:

[0133] In the present invention, due to the simplicity of the order crossover (OX operator) and the need not to perform conflict detection, it is selected as the crossover operator in the evolutionary process of the MFDE-ho algorithm. The crossover process of the OX operator is as follows:

[0134] First, select two parent generations, denoted as p1 and p2, and perform encoding on the parent generations. Then, randomly select the starting positions of crossover, denoted as s and e respectively, and retain the genes between the starting positions. Finally, find the other genes different from the retained fragments, and fill the different genes into the corresponding vacancies of their own offspring in the order of the other parent generation to generate new offspring.

[0135] The specific process of single-point mutation is as follows: First, randomly select a certain position as the mutation point a, and denote the gene at this mutation point as g1; Second, randomly generate a gene encoding within a certain range, denoted as g2; Then, find the position of gene g2 on the chromosome as the mutation point b; Finally, replace the gene at mutation point a with g2, and replace the gene at mutation point b with g1.

[0136] Task allocation scheme

[0137] Sample points are obtained by simulating and sampling the objective function through the Monte Carlo method, and then an auxiliary objective function is established by fitting the sample points with a Gaussian function. The genetic algorithm is used to perform iterative optimization on the objective function and the auxiliary objective function respectively;

[0138] Initialize the population according to the set algorithm parameters, and calculate the factor cost of this population under each iterative optimization task, that is, the objective function value;

[0139] After sorting the population list in ascending order of factor cost, mark the sequence index of the individuals in the population list, that is, the factor rank. When multiple individuals have the same factor cost, the random tie-breaking method (prior art in this field, refer to the non-patent literature "Random Tie-breaking with Stochastic Dominance") is adopted;

[0140] Calculate the scalar fitness of each individual according to the factor rank and determine the skill factor of this individual;

[0141] During the evolution process, use the crossover mutation operator, generate offspring through selective mating, and then evaluate the offspring population through selective imitation. After merging the parent population and the offspring population, sort according to the factor cost of the merged population, re-determine the factor rank, update the scalar fitness and skill factor of each individual in the merged population, and finally perform environmental selection on the individuals in the population through the elitist strategy to enter the next iteration until the evolution stop condition is met;

[0142] Output the task allocation result of multi-UAV energy guarantee, and obtain the allocation result as Figure 4 shown.

[0143] The chromosome coding sequence corresponding to this solution is: [12, 0, 17, 23, 22, 21, 2, 11, 20, 3, 0, 6, 0, 18, 5, 8, 9, 10, 7, 16, 0, 13, 14, 0, 4, 19, 0, 1, 0, 15]. The specific task allocation solution after decoding is as follows: UAV I1 executes tasks 12 and 17, UAV I2 executes tasks 23, 22, and 21 (i.e., only executes task T 15 ), UAV I3 executes tasks 2, 11, and 20, UAV I4 executes tasks 3 and 6, UAV I5 executes tasks 18 and 5; UAV I6 executes tasks 8, 9, and 10, UAV I7 executes tasks 7 and 16, UAV I8 executes tasks 13 and 14 (i.e., only executes task T9), UAV I9 executes tasks 4 and 19, UAV I 10 executes tasks 1 and 15.

[0144] Meanwhile, we depicted the convergence curve during the operation of the algorithm, as shown in Figure 5 .

[0145] Obviously, it can be concluded that the h-MFEA algorithm has good convergence in this application and can converge to the optimal result in a short time.

[0146] The beneficial effects of the present invention are as follows:

[0147] The present invention conducts research on multi-UAV task allocation, selects the battlefield individual soldier energy support scenario, establishes a task allocation model based on the actual application requirements, and obtains an effective task allocation solution through efficient algorithm solving.

[0148] As used herein, the term "preferred" is intended to be used as an example, illustration, or exemplification. Any aspect or design described herein as "preferred" should not necessarily be construed as more advantageous than other aspects or designs. Instead, the use of the term "preferred" is intended to present a concept in a specific manner. As used in this application, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" is intended to naturally include any one of the permutations. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.

[0149] Moreover, although the present disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art based on a reading and understanding of this specification and the drawings. The present disclosure includes all such modifications and variations and is limited only by the scope of the appended claims. In particular with respect to the various functions performed by the above-described components (e.g., elements, etc.), the terms used to describe such components are intended to correspond to any component that performs the specified function of the component (e.g., it is functionally equivalent), unless otherwise indicated, even if structurally different from the disclosed structure that performs the functions in the exemplary implementations of the present disclosure shown herein. In addition, although a particular feature of the present disclosure has been disclosed with respect to only one of several implementations, such feature may be combined with one or other features of other implementations as may be desired and advantageous for a given or particular application. Moreover, insofar as the terms "comprising," "having," "containing," or any variation thereof are used in a particular embodiment or claim, such terms are intended to be inclusive in a manner similar to the term "including."

[0150] Each functional unit in the embodiments of the present invention may be integrated into a processing module, may exist physically alone for each unit, or may be integrated into one module with two or more units. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a magnetic disk, an optical disk, or the like. Each of the above-mentioned devices or systems may execute the storage method in the corresponding method embodiment.

[0151] In summary, the above embodiments are one implementation manner of the present invention, but the implementation manners of the present invention are not limited by the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent substitution methods and are all included in the protection scope of the present invention.

Claims

1. A method for task allocation of multi-UAV energy guarantee, characterized in that, It includes the following steps: Multi-UAV mission assignment service analysis. A multi-UAV mission assignment scenario is described by a triple {I, T, C}. I = {I1, I2,..., I m} is the set of UAVs, indicating that there are m UAVs performing tasks in the battlefield. T = {T1, T2,... T n} is the set of tasks, indicating that there are n tasks to be executed in the battlefield. C represents the constraint conditions in the scenario; Establish UAV I i Objective function of the time required to complete all assigned tasks Use the Monte Carlo method to simulate and sample the target function to obtain sample points, and then use the Gaussian function to fit the sample points to establish an auxiliary target function. Use the genetic algorithm to perform iterative optimization on the target function and the auxiliary target function respectively; Initialize the population according to the set algorithm parameters, and calculate the factor cost of the population under each iterative optimization task, that is, the target function value; After sorting the population list in ascending order of factor cost, mark the sequence index of the individuals in the population list, that is, the factor level. When multiple individuals have the same factor cost, use the random tie-breaking method; Calculate the scalar fitness of each individual according to the factor level and determine the skill factor of the individual; During the evolution process, use the crossover and mutation operators, generate offspring through selective mating, and then evaluate the offspring population through selective imitation. After merging the parent population and the offspring population, sort according to the factor cost of the merged population, re-determine the factor level, update the scalar fitness and skill factors of each individual in the merged population, and finally perform environmental selection on the individuals in the population through the elite strategy to enter the next iteration until the evolution stop condition is met; Output the task allocation result of multi-UAV energy guarantee; 2. The multi-UAV energy guarantee task allocation method according to claim 1, characterized in that The UAV I i The time required to complete all the assigned tasks is as follows: The task assignment scheme is represented by two two-dimensional matrices A and B. Among them, matrix A is a natural number matrix, and the element a in matrix A ij represents the order in which UAV I i executes task T j . Matrix B is an integer matrix, and the element b in matrix B ij represents the number of batteries that UAV I i supports and guarantees to task point T j ; Calculate the total number of tasks that each UAV needs to execute from matrix A, denoted as count i ; Use to represent the task sequence that UAV I i needs to execute, where each element d ik represents the value of the k-th task that UAV I i is to execute, which is task F i represents the time required for UAV I i to complete all assigned tasks, including travel time and delivery time. The calculation formula of F i is as follows: Among them, represents the distance from the warehouse W to the mission i represents the distance from the mission i to the mission represents the distance from the warehouse W to the mission i is the delivery time of the drone I, V i i is the flight speed of the drone I i ​​​​​​​​ 3. The multi-UAV energy guarantee task allocation method according to claim 2, characterized in that There are the following constraint conditions: Unmanned aerial vehicle I i The load constraint during the mission execution, i.e.: is the maximum payload of the UAV I i ; The UAV meets the range limit constraint during the mission execution, that is: S i is the maximum flight range of the unmanned aerial vehicle I i ; The power supply request of each demand point must be met, so: Q j It is the number of battery cells required for the task point.

4. The multi-UAV energy guarantee task allocation method according to claim 3, characterized in that Establish the target function of the following target optimization model: minimize maxmize(F i ) In the current task allocation scenario, the goal of the model is to minimize the project duration of task completion, that is, the time spent by the last UAV to complete its task sequence, that is, the makespan. The target function is:

5. The multi-UAV energy guarantee task allocation method according to claim 4, wherein The Gaussian function is defined as follows: where d is the number of decision variables, a and b are parameters to be estimated, and x i represents the i-th variable, μ i and σ i are obtained by minimizing the error between the estimated value of the Gaussian function and the actual value of the sample; To fit the function y = f(x), x ∈ R through Gaussian process d , sample the original function to obtain n sampling points x = (x 1 , x 2 ,..., x n ) ∈ R d and their corresponding observed values y = (y 1 , y 2 ,..., y n ). For any two sampling points x i , x i ′ ∈ R d , the correlation c between them is defined as: Among them, 1 ≤ p i ≤ 2, which is used to measure the smoothness of the fitting function, and θ i ≥ 0 represents the importance of x i to the fitting function f(x). Then, the hyperparameter values of the Gaussian function are obtained by minimizing the error function, and the error function is defined as follows: Among them, C is an n×n matrix composed of c(x, x′), and I is an n×1 unit vector.

6. The multi-UAV energy guarantee task allocation method according to claim 5, characterized in that, Adopt the integer coding method, and use the task number as the chromosome gene; Each time a soldier requests energy supply, he will send information of a demand of 30 batteries to the control center to continue the electric energy required for the next 2 days of operation. A demand of 60 at a certain task point means that two soldiers have sent energy supply requests at this task point. Decompose this task into two subtasks with a demand of 30. In this way, all tasks in the task list are disassembled into multiple subtasks with a demand of 30, and each task point can only be visited once.

7. The multi-UAV energy guarantee mission allocation method according to claim 6, wherein Set up false tasks for supplementation to ensure that each UAV is assigned to multiple task points and facilitate subsequent genetic operations; the false tasks are numbered with negative integers and used as chromosome genes.

8. The multi-UAV energy guarantee task allocation method according to claim 7, characterized in that The mutation process of the genetic algorithm is as follows: First, randomly select a position as the mutation point a, and the gene at this mutation point is denoted as g1; second, randomly generate a gene encoding within a range, denoted as g2; then find the position of the gene g2 on the chromosome as the mutation point b; finally, replace the gene at the mutation point a with g2, and replace the gene at the mutation point b with g1.

9. The multi-UAV energy guarantee task allocation method according to claim 8, wherein Adopt the extraction technology to simplify the target function, that is, randomly select k components.

Images