An optimization strategy for multi-UAV collaborative task allocation based on improved wolf pack algorithm

Abstract

The present invention belongs to the field of drone control, and specifically relates to a multi-drone collaborative task allocation optimization strategy based on an improved wolf pack algorithm, including setting a proportional decision number that decreases with the number of iterations of the wolf pack algorithm, determining the number of scout wolves according to the proportional decision number and the total number of individual wolves; in the wandering stage, exchanging the corresponding codes of scout wolves with adjacent objective function values, if the objective function value becomes smaller after the code exchange, then updating the scout wolf code, otherwise not updating the scout wolf code; when the scout wolf objective function value exceeds the leader wolf objective function value, or reaches the maximum wandering number, entering the summoning stage of the wolf pack algorithm; and performing drone task allocation according to the optimal solution determined by the above wolf pack algorithm. The improved wolf pack algorithm proposed in the present invention can improve the convergence speed of the algorithm and the efficiency of task allocation for multi-drone collaborative tasks, and in the task allocation problem, the population is discretized to improve the robustness and applicability of task allocation.

Description

Technical Field

[0001] The present invention belongs to the field of unmanned aerial vehicle control, and in particular relates to a multi-unmanned aerial vehicle collaborative task allocation optimization strategy based on an improved wolf pack algorithm. Background Art

[0002] In the current UAV task allocation method based on IWPA (Improved Wolf Pack Algorithm), although the population update process moves towards the optimal solution, the update method is more inclined to the solution method of continuous problems, and still has a certain randomness and slow convergence speed, which affects the efficiency of task allocation. In addition, in the process of intelligently constraining and controlling the randomness of the wolf pack algorithm, it is not considered that the task allocation problem is a discrete problem and should have a specific discretization process. Therefore, the above existing methods have low robustness and applicability, and there is a certain gap between the optimal solution obtained by them and the actual optimal allocation strategy. Summary of the invention

[0003] In order to solve the above-mentioned problems in the prior art, that is, to improve the problem of existing UAV task allocation, the present invention provides a multi-UAV collaborative task allocation optimization strategy based on an improved wolf pack algorithm, including: encoding according to the UAV task requirements and task point numbers to obtain multiple individual wolves; setting an objective function according to the total cost of the UAV to execute the task; determining the distances of the corresponding individual wolves according to the difference between the objective functions of the individual wolves; selecting the individual wolf with the lowest objective function value as the head wolf; setting a proportional decision number that decreases with the number of iterations of the wolf pack algorithm, and determining the number of scout wolves according to the proportional decision number and the total number of individual wolves; and selecting a scout wolves according to the number of scout wolves. Take multiple individual wolves with lower objective function values ​​as scout wolves, and individual wolves other than the alpha wolf and the scout wolf as fierce wolves; in the wandering stage of the wolf pack algorithm, exchange the corresponding codes of the scout wolves with adjacent objective function values. If the objective function value becomes smaller after the code exchange, the scout wolf code is updated, otherwise the scout wolf code is not updated; when the objective function value of the scout wolf exceeds the objective function value of the alpha wolf, or the maximum number of wandering times is reached, enter the summoning stage of the wolf pack algorithm; in the summoning stage of the wolf pack algorithm, replace part of the code of the alpha wolf with the corresponding code of the fierce wolf. If the objective function value of the fierce wolf is less than the objective function value of the alpha wolf after the code replacement, and the distance between the alpha wolf and the fierce wolf is less than the threshold S min , then it enters the siege stage of the wolf pack algorithm; in the siege stage of the wolf pack algorithm, when updating the code, randomly replace the values ​​that meet the drone mission requirements, and randomly exchange the mission point vectors of the two individual wolves; according to the optimal solution determined by the above wolf pack algorithm, the drone task allocation is carried out.

[0004] Beneficial effects of the present invention:

[0005] The improved wolf pack algorithm proposed in the present invention can improve the convergence speed of the algorithm and the efficiency of task allocation for multi-UAV collaborative tasks. In the task allocation problem, the population is discretized to improve the robustness and applicability of task allocation.

[0006] Other features and advantages of the present invention will be described in the following description. The purpose and other advantages of the present invention can be realized and obtained through the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0008] Figure 1 The embodiment of the present invention provides a multi-UAV collaborative task allocation optimization strategy flow chart based on an improved wolf pack algorithm;

[0009] Figure 2 It is a schematic diagram of the outdoor ultra-high-definition live broadcast scene and task allocation assisted by drones;

[0010] Figure 3 is a flow chart of an improved WPA algorithm provided by an embodiment of the present invention;

[0011] Figure 4 This is a comparison chart of the effects of different task allocation algorithms;

[0012] Figure 5 It is the average objective function value curve corresponding to the change of distance-time weight ratio;

[0013] Figure 6 It is a comparison chart of the objective function convergence curve. DETAILED DESCRIPTION

[0014] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0015] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0016] With the help of advanced flight control technology and high-definition camera equipment, drones can quickly capture images and provide a variety of shooting angles, showing unique advantages in the field of high-definition live broadcasting. However, with the complexity of the high-definition live broadcast environment and the increase in demand, it is difficult for a single drone to complete complex task requirements in a timely manner. Multi-drone collaborative live broadcasting faces challenges such as mutual interference and inconsistent performance. Solving these problems requires not only the assistance of intelligent algorithms, but also consideration of constraints such as terrain and flight restrictions in real scenarios to meet the requirements of low latency and low energy consumption for ultra-high-definition live broadcasting, and to reasonably allocate tasks to different drones.

[0017] The embodiment of the present invention provides a multi-UAV collaborative task allocation optimization strategy based on an improved wolf pack algorithm, such as Figure 1 As shown, steps S1 to S6 are included:

[0018] S1: According to the UAV mission requirements and mission point numbers, multiple individual wolves are encoded; the objective function is set according to the total cost of the UAV mission; the distance between the corresponding individual wolves is determined according to the difference between the objective functions of the individual wolves; the individual wolf with the lowest objective function value is selected as the leader.

[0019] like Figure 2 In the drone-assisted outdoor ultra-high-definition live broadcast scene shown in the figure, all drones are located in the warehouse before performing the mission. After receiving the mission execution instruction, they set out from the warehouse and automatically return to the warehouse after completing the mission or running out of power. The drones can capture the wonderful scenes on the scene and transmit them back to the outdoor live broadcast vehicle through the 5G wireless network, and can also monitor various situations on the scene in real time.

[0020] Establish a heterogeneous UAV model. Assume that the UAV collection N is the maximum number of drones. The capacity of each drone includes the storage capacity C n and shooting ability S n Two. The outside broadcast van (OBV) serves as the base station in the intelligent live broadcast system, allocates bandwidth and channel resources to drones, and arranges drones to perform corresponding live broadcast tasks. When evaluating the heterogeneity of drones, the capability value of the drones can be used to determine the heterogeneity of the drones. n ,S n} to determine which mission it can perform. A drone with a specific capability value can perform only one mission or multiple different missions. Here, the different mission types mainly include ultra-high-definition live broadcast and real-time scene safety monitoring. Refer to the drone specifications and performance parameters provided by the manufacturer to obtain the performance parameters of the drone on different performance indicators.

[0021] In the task allocation problem, the population needs to be discretized to increase the applicability of the WPA algorithm. In view of the characteristics of heterogeneous UAVs, in order to clearly represent the task allocation scheme of the population individuals, the individual coding matrix X is introduced. m , where m represents the individual number of the population. m In the above example, each row represents a task requirement, and each column represents a task point, in the form of a column vector. In view of the characteristics of heterogeneous drones, Represents X m The 2D column vector corresponding to the i-th task point in , corresponds to the camera capture task and the site monitoring task. Since there is only one drone performing a task at a task point in the same period, two tasks cannot exist at the same time in the column vector. m ) is X m The objective function value is considered to be the concentration of prey odor perceived by the wolf pack. The distance between individual wolves like shown.

[0022] In some specific embodiments, encoding is performed according to the drone mission requirements, including: standardizing the storage performance parameters and shooting performance parameters of the drone to obtain a drone performance parameter matrix. Standardization here, also called normalization, refers to projecting all parameters into the interval (0,1) using a one-to-one mapping method such as the sigmod function. In order to ensure the smooth progress of high-definition live broadcast, in some specific embodiments, when the storage performance parameters or shooting performance parameters in the drone performance parameter matrix are less than 60%, the corresponding drone only meets the real-time scene safety monitoring task requirements; when the storage performance parameters and shooting performance parameters in the drone performance parameter matrix are not less than 60%, the corresponding drone meets the real-time scene safety monitoring task requirements and ultra-high-definition live broadcast task requirements.

[0023] Establish a task allocation model and define a task set Where 0 represents the warehouse location and I is the maximum number of task points. The distance between task points is calculated using the Euclidean distance formula dis(·), as follows: dis(i,j) = ||a i -a j ||, where a i Describe the position coordinates corresponding to the i-th task point. The total distance d that the n-th drone takes off from the warehouse to perform the task and then returns to the warehouse n Depend on Determine, where q n (i, j) is the decision parameter. When the nth UAV flies from task point i to task point j, the value of the decision parameter is 1. In other cases, q n (i,j) takes the value of 0. According to the decision parameter q of the nth UAVn (i,j) can get the sequence A of the tasks assigned to the drone n ={a0,a1,...,a k ,...,a K(n)}, where a0 represents the warehouse, a k (1≤k≤K(n)) represents the kth mission point, and K(n) represents the maximum number of missions performed by the nth drone. Then the flight time of the drone to the kth mission point is for: Where dis(i-1,i) represents the distance between task point i-1 and task point i, v avg Indicates the average flight speed of the drone. In addition to the drone flight time, the execution time of each task also needs to be considered. Specifically, the nth drone performs task a k The time required is assumed to be related to the capability value of the drone {C n ,S n} related, by t k (C n ,S n ) is expressed as follows. By adding the execution time of the task point to the flight time, we can get the waiting time t for the kth task to be executed. n,k , Among them, t i (C n ,S n ) indicates that the storage performance parameters and the shooting performance parameters are (C n ,S n ) is the time required for the nth UAV to perform the task corresponding to task point i; the total time t for all tasks of the nth UAV to wait for execution n for:

[0024] In some specific embodiments, the above objective function is specifically:

[0025]

[0026] C5:t total ≤t max

[0027] Where f represents the objective function value, and They are the proportional factors corresponding to the total flight distance of the UAV and the total waiting time for the task to be executed, satisfying and d total represents the total flight distance of all drones, t total represents the total waiting time for all tasks to be executed, N represents the total number of drones, and d maxrepresents the maximum distance each drone can fly, t max represents the maximum waiting time of all tasks, I represents the total number of task points, i and j represent the task point numbers, n represents the drone number, q n (i,j) represents the decision parameter. When the nth UAV flies from task point i to task point j, q n (i,j) is 1, otherwise q n (i,j) is 0, i or j is 0, indicating a drone warehouse, dis(i,j) indicates the distance between task point i and task point j, and C1 to C5 indicate the corresponding constraints.

[0028] In some specific embodiments, the total flight distance d of all drones is total Specifically N represents the total number of drones, n represents the drone number, d n It represents the total distance that the nth drone returns to the drone warehouse after starting from the drone warehouse to perform the mission; I represents the total number of mission points, i and j represent the mission point numbers, q n (i,j) represents the decision parameter. When the nth UAV flies from task point i to task point j, q n (i,j) is 1, otherwise q n (i,j) is 0, i or j is 0, indicating a drone warehouse, and dis(i,j) indicates the distance between task point i and task point j.

[0029] In some specific embodiments, the total waiting execution time of all tasks is t total Specifically t n Indicates the total time that all tasks of the nth drone are waiting to be executed; Where k represents the mission point number, K(n) represents the maximum number of missions performed by the nth UAV, and t n,k Indicates the waiting time for the task corresponding to the kth task point of the nth UAV to be executed; in, represents the flight time of the nth UAV to the kth mission point, t i (C n ,S n ) indicates that the storage performance parameters and the shooting performance parameters are (C n ,S n )’s nth UAV performs the task corresponding to task point i; Where dis(i-1,i) represents the distance between task point i-1 and task point i, v avg Indicates the average flight speed of the drone.

[0030] The size of the objective function value corresponds to the quality of the drone task allocation plan. The smaller the objective function value, the lower the consumption or cost of the drone to perform the task, and the higher the efficiency of task completion. In the constraints of the above objective function, constraint C1 limits each task point to only one drone to perform the task; constraint C2 sets the starting point and end point of the drone to be the warehouse; constraint C3 sets the total number of drones when the task point is used as the starting point to be equal to the total number of drones when the task point is used as the end point; constraint C4 limits the total flight distance of the drone to not exceed the maximum distance d that can be used for flight. max ; Constraint C5 limits the total waiting time for a task to be executed to no more than the maximum waiting time t for a task to be executed max .

[0031] During the entire iterative process of the wolf pack algorithm of this embodiment, the individual wolf with the lowest objective function value is selected as the leader.

[0032] In order to improve the algorithm convergence speed and task allocation efficiency, and improve the robustness and applicability of task allocation, this embodiment improves the traditional WPA (Wolf pack algorithm), and the specific improvements are shown in steps S2 to S5. The improved WPA algorithm process is as follows: Figure 3 shown. Figure 3 The fitness value in is also the objective function value.

[0033] S2: Set the proportional decision number that decreases with the number of iterations of the wolf pack algorithm, and determine the number of scout wolves based on the proportional decision number and the total number of individual wolves; based on the number of scout wolves, select multiple individual wolves with lower objective function values ​​as scout wolves, and individual wolves other than the alpha wolf and the scout wolf are fierce wolves.

[0034] Based on the individual wolf codes, the population performs wandering, calling and siege behaviors in sequence. For the first two behaviors, this embodiment introduces the crossover and mutation ideas in GA, and introduces a proportional control strategy into the WPA algorithm to optimize the process.

[0035] In some specific embodiments, the proportional decision number is specifically: Among them, p represents the proportional decision number, i* represents the current iteration number of the wolf pack algorithm, G max Represents the total number of iterations of the wolf pack algorithm.

[0036] In some specific embodiments, the number of wolves is scouted, specifically: Among them, m1 represents the number of scout wolves, M represents the maximum number of wolf packs, and p represents the proportional decision number.

[0037] According to the above-mentioned proportional decision number, some individuals with better objective function values ​​in the solution space except the alpha wolf are selected as scout wolves to update the population in the wandering stage. The scout wolves are similar to the scout wolves in the traditional WPA. In the traditional WPA, the wolf pack is divided into a fixed number of scout wolves and fierce wolves. As the number of iterations increases, the wolf pack gradually approaches the optimal solution. At this time, the fixed number of population updates will lead to a slower convergence speed and easily fall into the local optimum. Before entering the population grouping strategy, the proportional decision number is introduced and its value is dynamically adjusted. The population grouping strategy is the above-mentioned method of dividing multiple individual wolves into alpha wolves, scout wolves and fierce wolves.

[0038] Based on the above population proportion control strategy, in the early iteration stage, the value of m1 is small, and most of the individuals in the population perform calling behavior, which can quickly approach the potential optimal solution in the search space and accelerate the convergence of the algorithm; as the number of iterations increases, the value of m1 gradually increases. At this time, the population increases its own search range, which helps to escape the local optimum.

[0039] S3: In the wandering stage of the wolf pack algorithm, the corresponding codes of the scout wolves with adjacent objective function values ​​are exchanged. If the objective function value becomes smaller after the code exchange, the scout wolf code is updated, otherwise the scout wolf code is not updated; when the objective function value of the scout wolf exceeds the objective function value of the head wolf, or the maximum number of wandering times is reached, the wolf pack algorithm enters the summoning stage.

[0040] Specifically, the wandering behavior of traditional WPA is essentially a disordered and random exploration of the solution domain, which is easy to miss the possible optimal solution or fall into the local optimum. Therefore, in the individual wolf update process of the wolf pack wandering stage, GA (Genetic Algorithm) is combined to exchange the corresponding column vectors of two adjacent drone codes. The exchange rule is shown in the following formula.

[0041]

[0042] Represents X m The 2D column vector corresponding to the i-th task point in .

[0043] In order to increase the diversity of the solution space, the above process is repeated α times. Then the exchanged column elements are checked one by one, and the conflicting parts are randomly replaced until the performance requirements of the drone are met. The matrix after the exchange is X' m , if f(X' m )<f(X m ), then update the current exploration matrix, otherwise keep the original matrix unchanged. Repeat the above wandering behavior until f(X' m ) is greater than the optimal solution matrix X lead The corresponding objective function value f(X lead ), or the maximum number of walks J is reachedmax , then the system updates X lead And perform the summoning behavior, X lead Fingerwolf coding matrix.

[0044] S4: In the wolf pack algorithm summoning phase, the part of the code of the alpha wolf is replaced with the corresponding code of the fierce wolf. If the target function value of the fierce wolf after the replacement code is less than the target function value of the alpha wolf, and the distance between the alpha wolf and the fierce wolf is less than the threshold S min , then it enters the siege stage of the wolf pack algorithm.

[0045] Specifically, the calling behavior is initiated by the alpha wolf, and the remaining individual wolves, as a ferocious wolf pack (referred to as ferocious wolves), approach the alpha wolf with a larger step size, that is, the update of the individual will be based on the current optimal individual information. Since the calling behavior of traditional WPA is based on random selection, it cannot be ensured that each call can effectively guide the wolf pack to move in a better direction. In order to quickly find the optimal solution and speed up the convergence of the algorithm, the GA idea is introduced. By randomly copying part of the information in the current optimal individual, the corresponding position in the ferocious wolf pack individual is replaced. The copying method is shown in the following formula.

[0046]

[0047] Here, θ takes the value 0 or 1 with equal probability. Represents X m The 2D column vector corresponding to the i-th task point in . Represents X lead The 2D column vector corresponding to the i-th task point in .

[0048] After the swap, the matrix is ​​X" m , if f(X” m )>f(X lead ), the individual wolf replaces the alpha wolf and continues the calling behavior; otherwise, the fierce wolf continues to approach the alpha wolf until f(X lead )-f(X” m ) is less than a given threshold s min , and then turned to siege behavior.

[0049] S5: In the siege phase of the wolf pack algorithm, when updating the code, the values ​​that meet the drone mission requirements are randomly replaced, and the mission point vectors in two individual wolves are randomly exchanged.

[0050] Specifically, the joint siege behavior can be regarded as a small-scale search by all wolves around the optimal solution, that is, the position of the alpha wolf. Since the population is around the current optimal solution, a single individual update can easily cause the algorithm to fall into a local optimum. Therefore, this embodiment uses a combination of neighborhood operations within and between individuals to perform local search, which reduces the search range while increasing the ability to jump out of the local optimum.

[0051] Step 1: Introduce the mutation operation in GA into the matrix, that is, randomly determine the step c After a code bit that needs to be updated is found, the value of the corresponding column is randomly replaced until it meets the UAV capability requirements.

[0052] Step 2: Introduce a swap operator between individuals, that is, randomly select two different encoding matrices X m and X n , select a task point from each and Exchange the column vectors corresponding to the two task points.

[0053] The improved wolf pack algorithm of this embodiment is used to help escape from the local optimum and find a better strategy in the siege phase. After the iteration is completed, if the updated fitness (that is, the objective function value) value is smaller, the position matrix is ​​updated, otherwise the original matrix is ​​kept unchanged.

[0054] In the process of wolf pack updating, in order to maintain the overall number, it is necessary to introduce new individuals while continuously eliminating individuals with poor performance. Traditional WPA randomly generates coding matrices during updating, and the individuals generated lack competitiveness and cannot quickly converge to the optimal solution. In order to improve the replacement mechanism of the population, in some specific embodiments, the multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm also includes: when adding new individual wolves to the wolf pack, generating the new individual wolves includes: starting from the head wolf, in the order of the individual wolf objective function values ​​in the wolf pack from small to large, multiple individual wolves are selected into the reference group; according to the random replication probability, the individual wolves in the reference group are replicated to generate new individual wolves; the new individual wolves whose distance from the head wolf is less than the preset minimum wolf pack distance are discarded, and the individual wolves are regenerated.

[0055] In some specific embodiments, individual wolves in a reference group are replicated according to a random replication probability to generate new individual wolves, including: selecting codes at different positions of different individual wolves from the reference group, and copying the selected codes to corresponding positions of the new individual wolves according to the random replication probability.

[0056] Specifically, the top three individual wolves with the highest fitness values ​​in the iteration process are taken as the reference group X. r , randomly generated replication probability As shown in the following formula.

[0057]

[0058] Among them, X lead Refers to the wolf coding matrix, X1 refers to the individual wolf whose fitness value is ranked first from small to large except the wolf leader, and X2 refers to the individual wolf whose fitness value is ranked second from small to large except the wolf leader.

[0059] New individual wolf The value of selects different individual wolves in the reference group for information replication, which effectively increases the convergence speed of the algorithm and improves the initial competitiveness of the individuals, as shown in the following formula.

[0060]

[0061] Here, θ takes the value 0 or 1 with equal probability. Represents the new individual Wolf X new The 2D column vector corresponding to the i-th task point in . Represents the 2D column vector corresponding to the i-th task point in the encoding of the individual wolf copied by the new individual wolf.

[0062] The wolf pack generated by the above update strategy in the siege phase gathers around the position of the leader, making it difficult for the algorithm to jump out of the local optimum. In order to make further improvements, the concept of minimum wolf pack distance is introduced. The minimum wolf pack distance d min Refers to the distance d between the newly generated wolf pack individual m and the leader wolf when the population is replaced. m,lead The minimum value that should be satisfied. If the current individual d m,lead <d min , then it is necessary to regenerate the individual wolf coding matrix until the requirements are met.

[0063] S6: Allocate UAV tasks according to the optimal solution determined by the wolf pack algorithm.

[0064] When multiple drones perform task assignment, in order to minimize resource waste and improve task assignment efficiency, the drone's travel path should be arranged reasonably to ensure that it meets the constraints during the task execution. First, randomly select a task assignment plan, traverse the task points, and calculate the distance between the task points. Secondly, determine whether the current task point is a warehouse. If not, accumulate the flight distance d from the nth drone to the current kth task point. n,k and waiting time t n,k Then the allocation plan is judged whether it meets the constraints. If not, the process is terminated. If the current task point is a warehouse, then d n,k and t n,k The number of drones that are currently in use is cleared, and a backup drone is arranged to replace the current drone to perform the task. dis(i-1,i) represents the distance between task point i-1 and task point i. Finally, the task allocation scheme that satisfies the constraint conditions is output. After the constraint conditions of the individual matrix are screened, the IWPA provided by the present invention is iterated to find the global optimal solution of the population. The IWPA algorithm is executed multiple times under the same parameters, and the scheme with the smallest objective function value is selected as the optimal task allocation scheme of the system.

[0065] In this embodiment, the collaborative task allocation for multiple UAVs based on the improved wolf pack algorithm can improve the algorithm convergence speed and task allocation efficiency. In the task allocation problem, the population is discretized to improve the robustness and applicability of task allocation.

[0066] The technical effect of the multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm described in this embodiment is specifically analyzed as follows:

[0067] Assume that there are 3 drones and 10 target mission points in the simulation scene, among which drone 1 and drone 3 can perform 2 tasks, and drone 2 can only perform scene detection tasks. The location coordinates of drone mission points are generated randomly to increase the diversity and generalization of the model. Different task allocation algorithms are compared experimentally, and the specific results are as follows: Figure 4 As shown in the figure. Among them, the set of task points that need to be broadcasted in ultra-high definition is {i1,i4,i5,i7,i9,i 10}, the set of task points that require real-time scene security monitoring is {i2,i3,i6,i8}. Figure 4 The optimal solutions obtained by running the four algorithms 50 times independently are shown. Figure 4 The figure (a) in the upper left corner shows the flight path of the drone using the traditional WPA algorithm to assign tasks. Figure 4 Figure (b) in the upper right corner is the flight path of the drone using the ATARY algorithm to allocate tasks. ATARY is a mechanism based on the improved task allocation algorithm of Grey Wolf proposed in the literature (DDLieira, MSQuessada, LHVNakamura, S.Sampaio, REDe Grande and RIMeneguette,"Optimization ofthe Task Allocation Process in VEC with the GWO Bioinspired Algorithm,"202318th Iberian Conference on Information Systems and Technologies (CISTI), Aveiro, Portugal, 2023, pp. 1-7) to balance the number of smart terminals and data distribution in the network. This method adds virtual nodes to the original model of VRP to achieve a better task allocation solution. Figure 4Figure (c) in the lower left corner shows the flight path of a UAV using the TLFWA algorithm to allocate tasks. TLFWA is a novel improved fireworks algorithm (Tent-Levy Fireworks Algorithm, TLFWA) based on a discrete update process proposed in the literature (J.Yu, J.Guo, X.Zhang, C.Zhou, T.Xie and X.Han, "A Novel Tent-Levy Fireworks Algorithm for the UAV Task Allocation Problem Under UncertainEnvironment," in IEEE Access, vol.10, pp.102373-102385, 2022) for a multi-objective task allocation model with uncertainty. The algorithm combines integer coding, Tent chaotic mapping, and Levy variational strategy, introduces expected value criteria and chance constraints, and transforms the model into a deterministic optimization model, providing a new way to solve multi-objective and uncertain decision-making problems. Figure 4 The figure (d) in the lower right corner is the flight path of the UAV using the IWPA algorithm provided by the present invention to allocate tasks. Figure 4 It can be observed that the task paths of (b), (c), and (d) did not cross, while a path intersection occurred when the UAV in (a) performed the task, indicating that the anti-interference ability of the WPA algorithm is weak. All UAV tasks were completed within the constraints, and the simulation results matched the UAV performance parameters. It can be seen that the IWPA algorithm provided by the present invention is used to solve the problem of heterogeneous UAV task allocation, which can effectively reduce the interference between UAVs. Comparing the flight path lengths of the four algorithms, the total flight path of the UAV in (d) is shorter, so the UAV task execution efficiency using the IWPA algorithm provided by the present invention is higher.

[0068] The optimization goal of the wolf pack optimization algorithm is the optimal objective function value. In order to study the performance of the four algorithms when the distance-time weight ratio of the objective function is different, the average objective function value after the algorithm is run 50 times is taken for comparative analysis. The simulation results are as follows: Figure 5 shown. Figure 5 In , the lower the average objective function value, the better the task allocation effect of the optimal solution provided by the algorithm. Figure 5 From the data analysis in , it can be seen that the curve of the IWPA algorithm provided by the present invention is always lower than the curves of the other three algorithms at the same distance-time weight ratio, which shows that the IWPA algorithm provided by the present invention always has obvious advantages in finding the optimal objective function.

[0069] In order to test the convergence speed and stability of the IWPA algorithm provided by the present invention, Figure 6As shown in Figure 2, it is compared with the other three UAV task allocation algorithms. When is 1, the optimal solution convergence curves of the four algorithms running 50 times independently are as follows Figure 6 As shown. Figure 6 It can be seen that the objective function value curves of the four algorithms all show a rapid downward trend in the early stage of iteration, and this trend gradually slows down with the increase in the number of iterations. Among them, the IWPA provided by the present invention can quickly approach the potential optimal solution, and has the ability to quickly jump out of the local optimum. The objective function value converges first and stabilizes at a lower numerical level. The convergence speed of the other three algorithms is slow, and the objective function values ​​of traditional WPA and ATARY stabilize at a higher value after convergence. Traditional WPA performs poorly and is prone to fall into local optimality when the number of iterations is small. By comparison, the IWPA algorithm provided by the present invention has better convergence, the obtained objective function value is the smallest, and the task allocation effect is better.

[0070] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art can understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0071] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A multi-UAV collaborative task allocation optimization strategy based on an improved wolf pack algorithm, characterized in that: include: According to the UAV mission requirements and mission point numbers, multiple individual wolves are encoded; Set the objective function based on the total cost of the UAV’s mission execution; According to the difference between the objective functions of individual wolves, the distances corresponding to the individual wolves are determined; The individual wolf with the lowest objective function value is selected as the leader wolf; Set a proportional decision number that decreases with the number of iterations of the wolf pack algorithm, and determine the number of scout wolves based on the proportional decision number and the total number of individual wolves; based on the number of scout wolves, select multiple individual wolves with lower objective function values ​​as scout wolves, and individual wolves other than the alpha wolf and the scout wolf as fierce wolves; In the wandering phase of the wolf pack algorithm, the corresponding codes of the scout wolves with adjacent objective function values ​​are exchanged. If the objective function value becomes smaller after the code exchange, the scout wolf code is updated, otherwise the scout wolf code is not updated; When the scout wolf's objective function value exceeds that of the leader wolf, or reaches the maximum number of wandering times, the wolf pack algorithm enters the summoning phase; In the wolf pack algorithm summoning phase, the part of the code of the alpha wolf replaces the corresponding code of the fierce wolf. If the target function value of the fierce wolf after the replacement code is less than the target function value of the alpha wolf, and the distance between the alpha wolf and the fierce wolf is less than the threshold S min , then it enters the siege stage of the wolf pack algorithm; In the siege phase of the wolf pack algorithm, when updating the code, the values ​​that meet the drone mission requirements are randomly replaced, and the mission point vectors in two individual wolves are randomly exchanged; According to the optimal solution determined by the wolf pack algorithm, the drone tasks are allocated; in: When the storage performance parameters or shooting performance parameters in the drone performance parameter matrix are less than 60%, the corresponding drone only meets the real-time scene safety monitoring task requirements; When the storage performance parameters and shooting performance parameters in the drone performance parameter matrix are not less than 60%, the corresponding drone meets the real-time scene security monitoring task requirements and ultra-high-definition live broadcast task requirements; The objective function is specifically: C5:t total ≤t max Where f represents the objective function value, and They are the proportional factors corresponding to the total flight distance of the UAV and the total waiting time for the task to be executed, satisfying and d total Indicates the total flight distance of all drones. t total represents the total waiting time for all tasks to be executed, N represents the total number of drones, d max Indicates the maximum distance each drone can fly. t max represents the maximum waiting time of all tasks, I represents the total number of task points, i and j represent the task point numbers, n represents the drone number, q n (i,j) represents the decision parameter. When the nth UAV flies from task point i to task point j, q n (i,j) is 1, otherwise q n (i,j) is 0, i or j is 0, indicating a drone warehouse, dis(i,j) indicates the distance between task point i and task point j, and C1 to C5 indicate the corresponding constraints respectively; The total waiting time of all tasks is t total Specifically t n Indicates the total time that all tasks of the nth drone are waiting to be executed; Where k represents the mission point number, K(n) represents the maximum number of missions performed by the nth UAV, and t n,k Indicates the waiting time for the task corresponding to the kth task point of the nth UAV to be executed; in, represents the flight time of the nth UAV to the kth mission point, t i (C n ,S n ) indicates that the storage performance parameters and the shooting performance parameters are (C n ,S n )’s nth UAV performs the task corresponding to task point i; Where dis(i-1,i) represents the distance between task point i-1 and task point i, v avg Indicates the average flight speed of the drone.

2. The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm according to claim 1 is characterized in that: Encoding is performed according to the UAV mission requirements, including: The storage performance parameters and shooting performance parameters of the UAV are standardized to obtain the UAV performance parameter matrix.

3. The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm according to claim 1 is characterized in that: The total flight distance of all drones d total Specifically N represents the total number of drones, n represents the drone number, d n It represents the total distance that the nth drone returns to the drone warehouse after starting from the drone warehouse to perform the mission; I represents the total number of mission points, i and j represent the mission point numbers, q n (i,j) represents the decision parameter. When the nth UAV flies from task point i to task point j, q n (i,j) is 1, otherwise q n (i,j) is 0, i or j is 0, indicating a drone warehouse, and dis(i,j) indicates the distance between task point i and task point j.

4. The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm according to claim 1 is characterized in that: Proportional decision number, specifically: Among them, p represents the proportional decision number, i* represents the current iteration number of the wolf pack algorithm, G max Represents the total number of iterations of the wolf pack algorithm.

5. The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm according to claim 1 is characterized in that: Scout wolf populations, specifically: Among them, m1 represents the number of scout wolves, M represents the maximum number of wolf packs, and p represents the proportional decision number.

6. The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm according to claim 1 is characterized in that: The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm also includes: When a new individual wolf is added to the wolf pack, generating the new individual wolf includes: Starting from the alpha wolf, multiple individual wolves are selected into the reference group in the order of the objective function values ​​of the individual wolves in the wolf pack from small to large; According to the random replication probability, the individual wolves in the reference group are replicated to generate new individual wolves; New individual wolves whose distance to the leader wolf is less than the preset minimum wolf pack distance are discarded, and new individual wolves are generated.

7. The multi-UAV collaborative task allocation optimization strategy based on the improved wolf pack algorithm according to claim 6 is characterized in that: According to the random replication probability, the individual wolves in the reference group are replicated to generate new individual wolves, including: Select codes of different positions of different individual wolves from the reference group, and copy the selected codes to the corresponding positions of new individual wolves according to the random replication probability.

Images