Multi-UAV task allocation method based on multi-strategy fusion and improved dung beetle algorithm

Abstract

The present invention belongs to the field of UAV control technology, and specifically relates to a multi-UAV task allocation method based on a multi-strategy fusion improved dung beetle algorithm, which includes the following steps: using the RRT algorithm to solve the track and path distance between the UAV and the task point; using the quantitative objective function of minimizing the total cost of the UAV and maximizing the total benefit of the task to construct a task allocation mathematical model; setting the initialization parameters of the multi-strategy fusion improved dung beetle algorithm, initializing the position of the dung beetle population and differentiating the population; looping and iterating the position of the dung beetle population until the iteration end condition is met, and ending the population update; decoding the global optimal solution as the final optimal task allocation solution. The present invention can realize efficient, stable and real-time task scheduling in complex terrain and large-scale task environments, and meet the actual needs of large-scale UAV collaborative operations and complex task scheduling.

Description

Technical Field

[0001] The invention belongs to the technical field of unmanned aerial vehicle control, and in particular relates to a multi-unmanned aerial vehicle task allocation method based on a multi-strategy fusion improved dung beetle algorithm. Background Art

[0002] The Multi-UAV Task Allocation (MUTA) problem usually refers to the discussion of how to reasonably assign each task to each UAV to complete the entire task set given a set of tasks and a set of UAVs. This problem involves multiple key factors, including the performance parameters of the UAV, the urgency of the task, the priority of the task, and the coordination between UAVs. Specifically, the performance parameters of the UAV determine what types of tasks it is suitable for, the urgency of the task requires that some tasks must be completed first, and the priority of the task needs to be weighed under limited resources. In addition, coordination between UAVs is also crucial to ensure that they can work together efficiently and avoid conflicts and duplication of work. In general, the MUTA problem is a typical combinatorial optimization problem, whose goal is to find an optimal or suboptimal task allocation solution to maximize the efficiency of task execution and overall benefits while satisfying various constraints.

[0003] The Dung Beetle Optimizer (DBO) is a heuristic optimization algorithm inspired by the various behaviors of dung beetles in finding food, such as pushing a ball, dancing, foraging, stealing, and reproduction. It simulates these behaviors to find the optimal solution. Dung beetles use their sense of smell to find food sources and explore the optimal path in their surroundings. In the multi-UAV task allocation problem, the dung beetle algorithm has strong global search capabilities and good adaptability, and can be used as a potential global optimization algorithm. However, when dealing with complex multi-task and multi-UAV optimization problems, the standard dung beetle algorithm has shortcomings such as slow convergence and easy to fall into local optimality. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-UAV collaborative task allocation method based on a multi-strategy fusion improved dung beetle algorithm that can achieve efficient, stable and real-time task scheduling in complex terrain environments and large-scale task clusters.

[0005] The present invention adopts the following technical solution:

[0006] A multi-UAV collaborative task allocation method based on a multi-strategy fusion improved dung beetle algorithm comprises the following steps:

[0007] (1) Use the RRT algorithm to obtain the track and path distance between the UAV and the mission point;

[0008] (2) Based on the three-dimensional terrain environment and the flight distances between target points obtained in step (1), the objective function is quantified using two criteria: minimizing the total cost of the UAV and maximizing the total mission benefit, and then a mathematical model for task allocation is constructed;

[0009] (3) Set the initialization parameters of the multi-strategy fusion improved dung beetle algorithm, initialize the position of the dung beetle population and differentiate the population;

[0010] (4) Based on the improved dung beetle algorithm with multi-strategy fusion, the position of the dung beetle population is iterated cyclically, the local optimal solution obtained in each generation of population is recorded, and the global optimal solution is updated after the cycle ends; the population update ends when the iteration end condition is met;

[0011] (5) The globally optimal solution is decoded and used as the final optimal task allocation solution.

[0012] Furthermore, the method further includes step (6), evaluating the obtained task allocation scheme, checking the task execution time, load balancing and resource utilization indicators, and making adjustments based on the evaluation results.

[0013] Furthermore, in step (2), the penalty function strategy is introduced to transform the multi-UAV task collaborative allocation problem into an unconstrained optimization problem, and the objective function of the multi-UAV task collaborative allocation problem is defined as:

[0014]

[0015] in, is the total cost function for the UAV to perform the task, is the total revenue function for the drone to complete the task, is the discount factor of the penalty function term, is the penalty function term corresponding to the three constraints of the original objective function.

[0016] Furthermore, the total cost function of the drone's mission is As shown below:

[0017]

[0018] in, Indicates drone Execute the task The time delay function of For drones Execute the task actual time.

[0019] Furthermore, the total revenue function of the drone completing the task is As shown below:

[0020]

[0021] in, Representation task The time discount factor of Is a generated task The initial time, usually ; Is to obtain higher than the task The mission moment of gaining value in itself.

[0022] Furthermore, the penalty function The mathematical expression is:

[0023]

[0024]

[0025] in, is a generalized logical function, It is a parameter set according to the return distance of the drone. When the UAV is flying, it means the flight distance is less than its maximum range.

[0026] Furthermore, the penalty function The mathematical expression is:

[0027]

[0028] in, For drones Discount factor for resource carrying capacity.

[0029] Furthermore, the penalty function The mathematical expression is:

[0030]

[0031] in, For the task Discount factor for resource consumption.

[0032] Furthermore, in step (3), the initialization parameters of the multi-strategy fusion improved dung beetle algorithm include the population size , maximum number of iterations and the rolling ball dung beetle proportional constant based on the proportional decision number .

[0033] The rolling ball dung beetle proportional constant based on the proportional decision number Calculated by the following formula:

[0034]

[0035] in, The current iteration number of the dung beetle algorithm.

[0036] The formula for calculating the number of rolling dung beetles is:

[0037] .

[0038] Furthermore, in step (3), initializing the position of the dung beetle population and differentiating the population specifically includes the following steps:

[0039] (I) Randomly generate individual positions of the dung beetle population according to a uniform distribution :

[0040]

[0041] in, and To be the upper and lower bounds of the solution space, yes Random vectors with uniform distribution in the interval;

[0042] (II) Dynamic reverse learning is performed on the randomly generated initial population to obtain a new initial population, and then a greedy strategy is used to intensify the competition between the two populations to obtain the optimal initial population.

[0043] (III) Calculate the objective function value corresponding to the individuals in the initial population, and select the one with the lowest objective function value according to the number of rolling dung beetles. The individual is a rolling dung beetle, and then the individuals with the objective function value from low to high are classified into brooding dung beetles, small dung beetles, and thief dung beetles.

[0044] Furthermore, the specific process of step (4) is as follows:

[0045] (i) Rolling dung beetle position update

[0046] The probability of a dung beetle encountering an obstacle is assumed to be ,set up is a random number and .

[0047] when When it is in an obstacle-free state, the position update formula is as follows:

[0048] .

[0049] when When it is in the accessible state, generate a random number represents the deviation angle, and ,when The location is not updated when the position is set to zero, otherwise the location update is announced as follows:

[0050]

[0051] in, Indicates The objective function value in the generation population is less than The calculation formula for other random individuals with individual objective function values ​​is:

[0052]

[0053] in, Indicates The objective function value vector of the generation population, Indicates The first generation of the population Only the objective function value of the individual; Represents the global optimal position of the population.

[0054] (ii) Update of the position of brooding dung beetles

[0055] The boundaries of the area where female dung beetles were simulated to lay eggs were strictly limited as follows:

[0056]

[0057] in, and They represent the upper and lower limits of the spawning area respectively; Represents the local optimal position of the current iteration number; is a constant value that changes with the number of iterations, and Represents the maximum number of iterations.

[0058] The position of the brooding dung beetle also changes dynamically during the iteration process, and its position update formula is:

[0059]

[0060] in, Indicates The brood of dung beetles is Position information after iterations; and Two respectively dimensional independent random vectors, Represents the dimensionality of the optimization problem.

[0061] (iii) Location update of small dung beetles and thief dung beetles

[0062] The boundaries of the optimal foraging area for the small dung beetle are defined as follows:

[0063]

[0064] in, represents the global optimal position, and They represent the upper and lower bounds of the optimal foraging area. After introducing the water wave dynamic adaptive factor in this area, the position update formula of the foraging dung beetle is:

[0065]

[0066] in, represents random numbers that follow a normal distribution, Indicates belonging to A random vector of .

[0067] After introducing the water wave dynamic adaptive factor, the position update formula of the thief dung beetle is defined as:

[0068]

[0069] in, is a random vector of size 1×D that follows a normal distribution, Represents a constant.

[0070] (iv) Record the individual with the minimum objective function value in each generation of the population, which is the optimal solution for that generation of the population, and update the global optimal solution; when the iteration termination condition is met, the population update ends.

[0071] The beneficial effects of the present invention are:

[0072] (1) This invention builds a comprehensive task allocation model and defines the attributes of the task set and the UAV set in detail, including the task type, time requirement, task area, and the flight speed, load, and endurance of the UAV. These detailed attribute definitions enable the task allocation model to more accurately reflect the needs and limitations in actual applications, thereby providing a good foundation for the optimization of task allocation.

[0073] (2) The feasibility of the present invention proves that the combination of the dung beetle algorithm and the dynamic reverse learning strategy can be generalized and used in other swarm intelligence algorithms to improve the population initialization performance and effectively enhance the diversity of the initial population.

[0074] (3) During the algorithm iteration process, how to improve the algorithm's global search and local exploration capabilities is a key challenge. This paper introduces the global exploration strategy of the Osprey algorithm and the water wave dynamic adaptive factor to improve the algorithm update strategy, which comprehensively improves the population's global and local search capabilities;

[0075] (4) The present invention introduces a proportional decision number strategy to improve the dung beetle algorithm. In the early stage of the algorithm iteration, the number of dung beetles is small, and the population conducts a large-scale global search to accelerate the convergence of the algorithm. As the number of iterations increases, the number of dung beetles gradually increases. At this time, the individuals in the population increase their own search range, which helps to escape the local optimum.

[0076] In summary, the present invention significantly improves the efficiency and accuracy of multi-UAV task allocation problem by constructing a reasonable task allocation model and an innovative and improved multi-strategy fusion dung beetle algorithm. The multi-objective optimization design of the model enhances the rationality of task allocation and resource utilization, and the algorithm's multiple improvements in global and local search, dynamic fitness adjustment, etc., make the task allocation process more efficient and intelligent. It has good application prospects and significant practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 The present invention is a flowchart of the UAV task allocation based on the multi-strategy fusion improved dung beetle optimization algorithm.

[0078] Figure 2 This is a flowchart of the iterative part of the multi-strategy fusion improved dung beetle optimization algorithm of the present invention.

[0079] Figure 3 3D task scene diagram of an embodiment of the present invention, wherein (a) is a side view and (b) is a top view.

[0080] Figure 4 1 is an optimal task allocation sequence diagram of an embodiment of the present invention, wherein (a) is a side view and (b) is a top view.

[0081] Figure 5 Comparison diagram of objective function convergence curves of comparison algorithms. DETAILED DESCRIPTION

[0082] In order to make the present invention clearer and easier to understand, the following will describe in detail the implementation process of the multi-UAV task allocation method based on the improved dung beetle algorithm of the present invention through a specific embodiment in conjunction with the accompanying drawings. It can be understood that the specific embodiment described here is only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for the convenience of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0083] The purpose of the embodiment is to show how to implement the technical solution of the present invention in practical applications. Figure 1 As shown, the specific steps are as follows:

[0084] S1: An offline method is used to solve the track and flight distance between the starting point and the task point of the UAV through the RRT algorithm, and the flight distance between each point is used as one of the factors affecting task allocation.

[0085] S2: Establish a mathematical model for task allocation. Based on the three-dimensional terrain environment and the flight distance between task points obtained in step 1, the mathematical model for task allocation is constructed by quantifying the objective function using two criteria: minimizing the total cost of the UAV and maximizing the total benefit of completing the task.

[0086] The specific contents are as follows:

[0087] The drone set is described as: Assume for A collection of heterogeneous drones, the properties of each drone can be viewed as a cell array containing four elements: .

[0088] The task set is described as follows: yes randomly generated static ground task points, and the attributes of each task point are The meanings of the above parameters are shown in Table 1 below:

[0089] Table 1 Parameters included in the objective function and their meanings

[0090]

[0091] in: is the binary decision variable matrix for the task assignment problem, When indicating drone Execute the task Otherwise, drones Not performing tasks ; Represents the trajectory planning matrix between mission points, whose elements The mission point planned by the drone in step 1 With mission points The flight distance of the track.

[0092] Therefore, according to the characteristics of the multi-UAV collaborative task allocation model, it can be described as a total cost function of the UAVs: , the total revenue function of completing the task The combined optimization problem consisting of three constraints C1, C2, and C3, the mathematical expression of the objective function is:

[0093]

[0094] in, For drones Execute the task time.

[0095] 1) The total cost function for the drone to perform the task

[0096] The flight distance is taken as the main consideration of the cost function. The cost function ensures that the drone completes the task within the specified time while minimizing the flight distance cost of the drone; if the drone cannot complete the task within the specified time, the drone's return mission fails.

[0097] Based on the above discussion, its mathematical expression is as follows:

[0098]

[0099] in, Indicates drone Execute the task The time delay function of For drones Execute the task time.

[0100] 2) The total revenue function for the drone to complete the task is:

[0101] When completing an urgent task, you can get additional attack benefits; when completing an ordinary task, you can get an attack benefit that decays with time; when the task completion time exceeds the deadline, the benefit is always 0 regardless of whether the task is completed.

[0102] Based on the above discussion, its mathematical expression is as follows:

[0103]

[0104] in, Indicates the task The time discount factor of Is a generated task The initial time, usually ; Is to obtain higher than the task The mission moment of gaining value in itself.

[0105] On this basis, the penalty function strategy is introduced to transform the combinatorial optimization problem into an unconstrained optimization problem, so that the algorithm can fully consider the influence of constraints when finding the optimal solution and improve the accuracy of the task allocation model. The three constraints of the original objective function can be expressed as:

[0106] 1) Corresponding constraint C1

[0107] To ensure that the cumulative flight distance of each drone does not exceed its maximum endurance, the mathematical expression expressed as the penalty function term is:

[0108]

[0109]

[0110] in, is a generalized logical function, It is a parameter set according to the return distance of the drone. When , it means the flight distance of the drone is less than its maximum range; when It means that the drone can complete the assigned mission but cannot return home. At this time, the drone crashes on its own. Indicates that the drone is unable to complete all the tasks assigned to it.

[0111] 2) Corresponding constraint C2

[0112] To ensure that the resources consumed by the drone when performing a mission do not exceed the maximum amount of resources it can carry, the mathematical expression expressed as the penalty function term is:

[0113]

[0114] in, For drones Discount factor for resource carrying capacity.

[0115] 3) Corresponding constraint C3

[0116] Ensure that the resources allocated to a task are greater than the resources required to complete the task. The mathematical expression expressed as the penalty function term is:

[0117]

[0118] in, For the task Discount factor for resource consumption.

[0119] Using the above three penalty functions, the multi-UAV task collaborative allocation problem is transformed into an unconstrained optimization problem. Therefore, the objective function of the multi-UAV collaborative task allocation problem is defined as:

[0120]

[0121] in, is the discount factor of the penalty function term.

[0122] The present invention improves the traditional DBO algorithm by integrating multiple strategies to improve the algorithm convergence speed and the quality of the solution, and improve the wide applicability of the algorithm. The specific improvements are shown in steps S3 to S7. The improved DBO algorithm flow is detailed in Figure 2 .

[0123] S3: Set the initialization parameters of the dung beetle algorithm

[0124] The specific sub-steps are as follows:

[0125] S3.1: Define the population size as , the maximum number of iterations is , the proportion constant of the whole population of dung beetles is .

[0126] The fixed number of population updates will lead to a slow convergence speed and easy to fall into the local optimum. Therefore, based on the original algorithm parameters, the present invention introduces a proportional decision number strategy. It is defined as the proportional decision number that decreases with the current number of iterations of the algorithm, and its value is adjusted dynamically. The specific calculation formula is:

[0127]

[0128] in, is the current iteration number of the dung beetle algorithm; the calculation formula for the number of rolling dung beetles is:

[0129]

[0130] Based on the above population ratio control strategy, in the early iteration stage The value of is small, most of the population is foraging, and the potential optimal solution can be quickly approached in the search space, which speeds up the convergence of the algorithm. As the number of iterations increases, The value of gradually increases. At this time, the population increases its own search range, which helps to jump out of the local optimum.

[0131] S3.2: Initialize the individual positions of the dung beetle population.

[0132] First, the individual positions of the dung beetle population are randomly generated according to a uniform distribution. :

[0133]

[0134] Each individual position of a dung beetle represents a feasible solution in the solution space, where and To be the upper and lower bounds of the solution space, yes The random vector with uniform distribution in the interval can further be expressed as the following matrix form:

[0135] .

[0136] S3.3: Dynamic reverse learning strategy.

[0137] In order to improve the quality of the initial population, the present invention introduces a dynamic reverse learning strategy, performs reverse processing on the randomly generated initial population to obtain a new initial population, and then uses a greedy strategy to intensify the competition between the two populations to obtain the best initial solution. Therefore, the dynamic reverse learning strategy is used to enhance the quality of the initial population in the search space, and the method is as follows:

[0138]

[0139] In the formula, Represents the initialized population randomly generated by the DBO algorithm; , Between 0 and 1; first Perform dynamic reverse learning to obtain a new population ; then and Merge into ;calculate The objective function value is calculated by using a greedy strategy to intensify inter-species competition, thereby obtaining the required optimal initialization population. This method is used to speed up the algorithm's approach to the optimal solution, thereby improving the algorithm's convergence speed.

[0140] S3.4: Calculate the objective function value corresponding to the individuals in the initial population, and select the one with the lowest objective function value according to the number of rolling dung beetles. The individual is a rolling dung beetle, and then the individuals with the objective function value from low to high are classified as brooding dung beetles, small dung beetles, and thief dung beetles.

[0141] S4: Population cycle iterative update

[0142] Based on the above dung beetle population initialization and classification, the improved algorithm update strategy is used to update the individual positions. The specific implementation steps are as follows:

[0143] S4.1: Update of the position of the rolling dung beetle.

[0144] Inspired by the strategy of osprey optimization algorithm in which osprey populations share position information and cooperate in searching for prey during the exploration phase, a global exploration strategy is proposed to guide dung beetles to update their positions, so as to improve the lack of global search ability of standard DBO.

[0145] The specific update strategy is as follows:

[0146] The position update of the dung beetle is divided into two cases: with obstacles and without obstacles. In this strategy, the probability of the dung beetle encountering an obstacle is assumed to be ,set up is a random number and ,when When it is in an obstacle-free state, the position update formula is as follows:

[0147]

[0148] Random Numbers When it is in the accessible state, generate a random number represents the deviation angle, and ,when The position is not updated when , otherwise the position update formula is as follows:

[0149]

[0150] in, Indicates The objective function value in the generation population is less than The objective function value of each individual is calculated as follows:

[0151]

[0152] in, Indicates The objective function value vector of the generation population, Indicates The first generation of the population The objective function value of the individual; Represents the global optimal position of the population; other parameters are defined the same as the traditional DBO algorithm.

[0153] After applying the above strategy for improvement, individuals within the dung beetle can not only use their own information, but also make full use of the information shared by other dung beetle individuals to iterate their positions. This improvement not only greatly improves the efficiency of global exploration, but also helps the algorithm to effectively jump out of the local optimum, laying a solid foundation for other types of subpopulations to perform precise optimization in local space.

[0154] S4.2: Update on the location of brooding dung beetles.

[0155] In nature, dung beetles will roll their dung balls to a relatively safe place and use that area as a place to lay their eggs. Based on the above behavior, DBO proposed a boundary selection strategy to simulate the area where female dung beetles lay their eggs. The boundaries of this area are strictly limited as follows:

[0156]

[0157] in, and They represent the upper and lower limits of the spawning area respectively; Represents the local optimal position of the current iteration number; is a constant value that changes with the number of iterations, and Represents the maximum number of iterations.

[0158] The position of the brooding dung beetle also changes dynamically during the iteration process, and its position update formula is:

[0159]

[0160] in, Indicates The brood of dung beetles is Position information after iterations; and Two respectively dimensional independent random vectors, Represents the dimensionality of the optimization problem.

[0161] S4.3: Update of the positions of the small dung beetle and the thief dung beetle.

[0162] In order to enhance the local search capability of the algorithm, the present invention introduces a water wave dynamic adaptive factor in the position update of the small dung beetle and the small thief dung beetle, and uses the dynamic change uncertainty of the water wave to increase the search range of the two types of dung beetles, so as to avoid the blind following of the dung beetle population and cause the algorithm to fall into the local optimum. It is described as follows:

[0163] The water wave dynamic adaptive factors are as follows:

[0164]

[0165] Where t is the current iteration number and T is the maximum iteration number.

[0166] First, the boundaries of the optimal foraging area for the dung beetle are defined as follows:

[0167]

[0168] in, represents the global optimal position, and Respectively represent the upper and lower bounds of the optimal foraging area. After introducing the water wave dynamic adaptive factor in this area, the position update formula of the foraging dung beetle is:

[0169]

[0170] in, represents random numbers that follow a normal distribution, Indicates belonging to A random vector of .

[0171] After introducing the water wave dynamic adaptive factor, the position update formula of the thief dung beetle is defined as:

[0172]

[0173] in, is a random vector of size 1×D that follows a normal distribution, Represents a constant.

[0174] S5: Based on the above position update formula, the dung beetle population is iterated cyclically to solve the task allocation problem. The individual with the minimum objective function value in each generation of population is recorded as the optimal solution for this generation of population, and the global optimal solution is updated. When the iteration termination condition is met, the population update is terminated.

[0175] S6: Decode the global optimal solution as the final optimal task allocation solution.

[0176] S7: Evaluate the obtained task allocation plan, check indicators such as task execution time, load balancing and resource utilization, and make necessary adjustments based on the evaluation results to ensure that the task can be completed efficiently and accurately.

[0177] Calculation Example

[0178] The simulation scene is set in a three-dimensional space area G, which contains 10 peaks, which are set as no-fly zones. In addition, there is a drone base and 10 enemy target mission points in the scene. The side view and top view are as follows: Figure 3 As shown in Table 2, there are three types of drones in the drone base. The specific parameters are shown in Table 2. The position coordinates of the enemy target mission point are randomly generated within the mission scenario to improve the diversity and generalization ability of the model. The parameters generated in this example are shown in Table 3. Based on the above settings, the RRT algorithm is used in step 1 to calculate the distance from the drone base to each mission point and the distance between the mission points. The results are shown in Table 4 below. The total number of iterations of the optimization model in this example is , the initial population size .

[0179] Table 2 Parameters of different types of drones

[0180]

[0181] Table 3 Target mission point parameters

[0182]

[0183] Table 4 Distance between drone base and target point

[0184]

[0185] Based on the above task scenarios and parameter settings, the optimal task allocation sequence and task completion time window are calculated. The specific information is shown in Table 5 and Table 6. At the same time, in order to more intuitively display the results of task allocation, the task allocation result diagram is shown in Figure 4 shown.

[0186] Table 5 Task allocation scheme sequence

[0187]

[0188] Table 6 Task completion time window

[0189]

[0190] In order to further demonstrate the superiority of the algorithm of the present invention, the present invention has conducted simulation comparison experiments of multiple algorithms. The MSDBO algorithm of the present invention is compared and analyzed with the Dung Beetle Optimizer (DBO), Harris Hawks Optimization (HHO), and Seagull Optimization Algorithm (SOA).

[0191] 50 independent simulation experiments were conducted, and the total track cost of the UAV to complete the overall task was recorded as shown in Table 7. According to the data calculation, the task allocation scheme obtained by the MSDBO algorithm proposed in the present invention has the smallest total track cost of the overall task, which is 10.28% lower than the original DBO algorithm, while the HHO and SOA algorithms are 4.87% and 3.22% lower than the DBO algorithm, respectively. The improved MSDBO algorithm can significantly reduce the resource consumption of the UAV and the overall task completion time, greatly improve the quality of the obtained objective function value, and thus make the task allocation scheme more efficient and reasonable; the comparison chart of the objective function convergence curve can be obtained by recording the data of each iteration, as shown in the figure. Figure 5 As shown in the figure, it can be clearly seen that the improved algorithm performs well in both convergence speed and optimization effect, and the resulting allocation sequence successfully converges to the global optimum. These simulation experimental results strongly confirm that the multi-UAV collaborative task allocation method based on the improved dung beetle optimization algorithm can not only achieve efficient task allocation, but also show strong optimization capabilities in various complex scenarios, which is significantly better than traditional algorithms.

[0192] Table 7 Total track cost of the UAV to complete the overall mission

[0193]

[0194] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-UAV task allocation method based on multi-strategy fusion improved dung beetle algorithm, characterized in that: It includes the following steps: (1) Use the RRT algorithm to solve the track and path distance between the UAV and the mission point; (2) Based on the three-dimensional terrain environment and the flight distance between each target point obtained in step (1), the objective function is quantified by minimizing the total cost of the UAV and maximizing the total mission benefit, and then a mathematical model for task allocation is constructed; The penalty function strategy is introduced to transform the multi-UAV task collaborative allocation problem into an unconstrained optimization problem. The objective function of the multi-UAV collaborative task allocation problem is defined as: F(X,S)=f p (X,S)-f r (t ij )+ω(p a (S)+p b (X)+p c (X)) Among them, f p (X,S) is the total cost function of the UAV to perform the task, f r (t ij ) is the total revenue function of the UAV to complete the task, ω is the discount factor of the penalty function term, p a (S)+p b (X)+p c (X) is the penalty function term corresponding to the three constraints of the original objective function; The total cost function f of the drone's mission p (X,S) is shown in the following formula: in, represents the time delay function of UAV i executing task j; t ij is the actual time for UAV i to perform task j, s jk is the path distance between UAV i from task j to task k; (3) Setting the initialization parameters of the multi-strategy fusion improved dung beetle algorithm, initializing the position of the dung beetle population and differentiating the population; (4) Based on the improved dung beetle algorithm with multi-strategy fusion, the position of the dung beetle population is iterated cyclically, the local optimal solution obtained in the previous generations of populations is recorded, and the global optimal solution is updated after the cycle ends; when the iteration end condition is met, the population update ends; (5) The global optimal solution is decoded and used as the final optimal task allocation solution.

2. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 1 is characterized in that: The total profit function f of the drone completing the task r (t ij ) is shown in the following formula: Among them, λ j >0 indicates the time discount factor of task j; is the initial time to generate task j, usually It is the task moment that obtains a benefit higher than the value of task j itself.

3. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 2 is characterized in that: The penalty function term p a The mathematical expression of (S) is: Where: l(β) is the generalized logic function, α i is a parameter set according to the return distance of the UAV; when l(·) = 0, it means that the flight distance of the UAV is less than its maximum range.

4. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 3 is characterized in that: The penalty function term p b The mathematical expression of (X) is: Among them, pb i is the discount factor for the amount of resources carried by UAV i.

5. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 4 is characterized in that: The penalty function term p c The mathematical expression of (X) is: Among them, pc j is the discount factor of the resource consumption of task j.

6. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 5 is characterized in that: In step (3), the initialization parameters of the multi-strategy fusion improved dung beetle algorithm include the population size N, the maximum number of iterations T, and the rolling ball dung beetle proportional constant P based on the proportional decision number; The rolling ball dung beetle proportional constant P based on the proportional decision number is calculated by the following formula: P=0.1+0.9e -10i / T Among them, i is the current iteration number of the dung beetle algorithm; The formula for calculating the number of rolling dung beetles is:

7. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 6 is characterized in that: In step (3), initializing the position of the dung beetle population and differentiating the population specifically includes the following steps: (I) Randomly generate individual positions X of the dung beetle population according to uniform distribution i : X i =L+r×(UB-LB)1≤i≤N Among them, UB and LB are the upper and lower bounds of the solution space, and r is a random vector uniformly distributed in the interval [0,1]; (II) Dynamic reverse learning is performed on the randomly generated initial population to obtain a new initial population, and then a greedy strategy is used to intensify the competition between the two populations to obtain the best initial solution; (III) According to the number of ball-rolling dung beetles, select the n1 individuals with the lowest objective function value as ball-rolling dung beetles, and then classify the individuals with the objective function values ​​from low to high into brooding dung beetles, small dung beetles, and thief dung beetles.

8. The multi-UAV task allocation method based on the multi-strategy fusion improved dung beetle algorithm according to claim 7 is characterized in that: The specific process of step (4) is as follows: (i) Rolling dung beetle position update: The probability of a dung beetle encountering an obstacle is assumed to be γ, and λ is a random number and λ∈[0,1], When λ<γ, it is in an obstacle-free state, and the position update formula is as follows: x i (t+1)=x rand +α×k×x i (t-1)+b×△x △x=|x i (t)-X w | When λ≥γ, it is in an obstacle-free state. A random number θ is generated to represent the deviation angle, and θ∈[0,π]. When θ=0,π / 2,π, the position is not updated. Otherwise, the position update is shown as follows: x i (t+1)=x rand +tan(θ)|x i (t)-x i (t-1)| Among them, x rand It represents other random individuals in the t-th generation population whose objective function value is less than that of the i-th individual. The calculation formula is: Among them, F t represents the objective function value vector of the t-th generation population, F t i represents the objective function value of the i-th individual in the t-th generation population; X b Represents the global optimal position of the population; (ii) Update of the position of brooding dung beetles The boundaries of the area where female dung beetles were simulated to lay eggs were strictly limited as follows: Among them, UB * With LB * Respectively represent the upper and lower limits of the spawning area; X * Represents the local optimal position of the current iteration number; R is a constant value that changes with the number of iterations, R = 1-t / T and T represents the maximum number of iterations; The position of the brooding dung beetle also changes dynamically during the iteration process, and its position update formula is: B i (t+1)=X * +b1×(B i (t)-LB * )+b2×(B i (t)-UB * ) Among them, B i (t) represents the position information of the i-th brooding dung beetle after the t-th iteration; b1 and b2 are two d-dimensional independent random vectors, d represents the dimension of the optimization problem, (iii) Update of the location of the small dung beetle and the thief dung beetle The boundaries of the optimal foraging area for the small dung beetle are defined as follows: Among them, X b represents the global optimal position, UB' and LB' represent the upper and lower bounds of the optimal foraging area, respectively. After introducing the water wave dynamic adaptive factor in this area, the position update formula of the foraging dung beetle is: x i (t+1)=x i (t)+C1×(λ·x i (t)-LB')+C2×(λ·x i (t)-UB') Among them, C1 represents a random number that obeys the normal distribution, and C2 represents a random vector belonging to (0,1); After introducing the water wave dynamic adaptive factor, the position update formula of the thief dung beetle is defined as: x i (t+1)=X b +S×g×(|λ·x i (t)-X * |+|λ·x i (t)-X b |) Where g is a random vector of size 1×D that follows a normal distribution, and S represents a constant; (iv) Record the individual with the minimum fitness value in each generation of the population, which is the optimal solution for this generation of the population, and update the global optimal solution; when the iteration termination condition is met, the population update ends.

Images