Multi-unmanned aerial vehicle task collaborative planning method and system based on comprehensive cost constraint

Abstract

The invention discloses a multi-unmanned aerial vehicle task collaborative planning method and system based on comprehensive cost constraint, and belongs to the technical field of unmanned aerial vehicle system design and application, and the method comprises the steps: carrying out the data initialization in an image of a target region, and determining task point information and unmanned aerial vehicle information in the image; determining a distance cost and a path cost based on task point information and unmanned aerial vehicle information; then combining the path cost and the distance cost to obtain a fitness function of the comprehensive cost; task allocation is carried out on the unmanned aerial vehicle through a fitness function based on a genetic algorithm, and an optimal task allocation result is obtained; and finally, performing sequence optimization on the optimal task allocation result based on a grey wolf algorithm to complete collaborative planning of the tasks of the multiple unmanned aerial vehicles. According to the method, for the multi-unmanned aerial vehicle collaborative planning problem, the path cost value is predicted, the task is allocated to each unmanned aerial vehicle through the genetic algorithm, the optimal task allocation is obtained by adopting the fitness function of the comprehensive cost, the total operation cost reaches the lowest, and the working efficiency is higher.

Description

Technical Field

[0001] This application belongs to the technical field of unmanned aerial vehicle system design and application, and specifically relates to a multi-UAV task collaborative planning method and system based on comprehensive cost constraints. Background Art

[0002] The task allocation problem refers to allocating corresponding task objectives to each robot and determining the execution order between each objective under the constraints of the environment and task resources to ensure the optimization of the task execution result. Multi-UAV task allocation is an important link to ensure the execution of multi-UAV tasks and is also a prerequisite for the robot to perform path planning. The coupling of multi-UAV collaborative path planning and task allocation problems is mainly manifested in several aspects such as multi-constraint optimization coupling, interactive coupling, and collaborative planning coupling. Among them, interactive coupling comprehensively considers the coordination of task requirements, priority factors, etc., making the UAV system more efficient in completing all tasks. For its decoupling strategy, there are currently mainly three types: multi-constraint decoupling strategy and modular decoupling strategy. Multi-constraint decoupling is mostly a numerical optimization method, and its constraint conditions include the shortest path, the lowest energy consumption, and motion constraints. The optimal solution of the multi-constraint problem is obtained through algorithms such as GA (Genetic Algorithm) and SCP (Sequential Convex Programming), but the problem complexity has not been reduced and the computational workload is large; modular decoupling first predicts the path, allocates tasks according to the predicted path cost, and uses a path planning algorithm to plan the path for each UAV's task. This decoupling framework achieves the effect of hierarchical decoupling, reducing the difficulty of multi-UAV collaborative planning problems, but it is difficult to ensure that all constraint conditions are met in actual operation.

[0003] It can be seen that although remarkable achievements have been made in the multi-UAV collaborative planning problem, it is still difficult to cope with complex real environments. How the multi-UAV system effectively coordinates each optimization objective in complex task scenarios to achieve the optimal overall performance of the system has certain research value. Summary of the Invention

[0004] Object of the Invention: This application develops a multi-UAV task collaborative planning method and system based on comprehensive cost constraints, aiming to solve the technical problems in the prior art.

[0005] Technical Solution: In the first aspect, an embodiment of this application provides a multi-UAV task collaborative planning method based on comprehensive cost constraints, including:

[0006] Obtain an image of the target area, and determine the information of task points and UAVs in the image;

[0007] Obtain the distance cost from the UAV to the mission point, and the path cost of the UAV between the mission points;

[0008] Based on the path cost and the distance cost, obtain the fitness function of the comprehensive cost;

[0009] Based on the genetic algorithm and combined with the fitness function, perform task allocation for the UAV to obtain the optimal task allocation;

[0010] Based on the grey wolf algorithm, optimize the sequence of the optimal task allocation to complete the multi-UAV task collaborative planning.

[0011] In some embodiments, the steps of obtaining the path cost include:

[0012] Determine the starting point and the ending point, and establish search trees at both the starting point and the ending point;

[0013] Determine the expansion tree and the connection tree in the two search trees;

[0014] Randomly select sampling points in the image through the rapidly-exploring random tree connection algorithm as new nodes;

[0015] In the expansion tree, backtrack each node, and obtain the best parent node of the new node based on the cross-product collision detection algorithm;

[0016] Obtain the direct connection path between the node closest to the new node in the connection tree and the expansion tree as the initial path;

[0017] Optimize the initial path to obtain the final path, and obtain the path cost based on the final path.

[0018] In some embodiments, the steps of obtaining the best parent node of the new node based on the cross-product collision detection algorithm include:

[0019] Determine the vector formed by the obstacle edge line segments as the obstacle edge line segment vector;

[0020] Determine the vector formed by the connection line between the new node and the nodes in the expansion tree as the path line segment vector;

[0021] Obtain the cross product of the obstacle edge line segment vector and the path line segment vector;

[0022] In response to the cross product being greater than zero, determine that there is no collision interference between the obstacle edge line segment vector and the path line segment vector;

[0023] In response to the cross product being equal to zero and the cross product of the cross vectors being zero, determine that there is no collision interference between the obstacle edge line segment vector and the path line segment vector;

[0024] Determine the node in the extended tree corresponding to the minimum path segment vector without collision interference as the optimal parent node of the new node.

[0025] In some embodiments, the step of optimizing the initial path includes:

[0026] Among the nodes of the initial path, obtain a first node, a second node, and a third node;

[0027] Connect the second node and the first node to form a first edge, and connect the second node and the third node to form a second edge;

[0028] Determine a fourth node on the first edge and a fifth node on the second edge, where the distances from the fourth node and the fifth node to the second node are the same;

[0029] Connect the fourth node and the fifth node to obtain a third edge;

[0030] In response to the third edge having no collision interference with the obstacle, update the fourth node to the first node and update the fifth node to the third node.

[0031] In some embodiments, the step of optimizing the initial path includes:

[0032] Among the nodes of the initial path, obtain a first node, a second node, and a third node;

[0033] Connect the second node and the first node to form a first edge, and connect the second node and the third node to form a second edge;

[0034] Determine a fourth node on the first edge and a fifth node on the second edge;

[0035] Connect the fourth node and the second node to form a third edge, and connect the fifth node and the second node to form a fourth edge. The ratio of the length of the third edge to the length of the first edge is a, and the ratio of the length of the fourth edge to the length of the second edge is b, satisfying: a = b;

[0036] Connect the fourth node and the fifth node to obtain a fifth edge;

[0037] In response to the fifth edge having no collision interference with the obstacle, update the fourth node to the first node and update the fifth node to the third node.

[0038] In some embodiments, the step of performing task allocation on the UAV based on a genetic algorithm and in combination with the fitness function to obtain an optimal task allocation includes:

[0039] Initialize the population information based on the genetic algorithm, combining the data of the unmanned aerial vehicle and the data of the task points, and construct the gene sequence based on the population information;

[0040] Obtain the fitness of the chromosomes in the gene sequence based on the fitness function of the comprehensive cost;

[0041] Select the parent chromosomes in the gene sequence based on the fitness;

[0042] Perform crossover based on the parent chromosomes to obtain the offspring gene sequence;

[0043] Perform mutation operations on the chromosomes in the offspring gene sequence based on the reverse mutation method to obtain a new generation of population;

[0044] In response to the new generation of population meeting the population threshold, determine the optimal task allocation based on the new generation of population.

[0045] In some embodiments, the representation formula of the gene sequence includes:

[0046]

[0047] where pop is the gene sequence; k is the number of chromosomes, used to represent the number of preset solutions; n is the total number of tasks; a kn is the task point serial number.

[0048] In some embodiments, the step of performing crossover based on the parent chromosomes to obtain the offspring gene sequence includes:

[0049] Determine two crossover points in the parent chromosomes based on the partially matched crossover method;

[0050] Determine the crossover region based on the two crossover points;

[0051] Exchange the crossover region and determine the mapping relationship to obtain the offspring gene sequence.

[0052] In some embodiments, the step of performing sequence optimization on the optimal task allocation based on the grey wolf algorithm includes:

[0053] Determine that the preset solution for the sequence optimization is the position data of the grey wolf;

[0054] Determine that the number of preset solutions for the sequence optimization is the population quantity of the grey wolf;

[0055] Determine the hunting model and the pursuit model based on the position data and the population quantity, where the representation formula of the hunting model includes:

[0056] D = |C·X p (t)-X(t)|;

[0057] X(t + 1) = X p (t) - A·D;

[0058] Wherein, X p is the position vector of the prey; X is the position vector of the gray wolf, D represents the distance between the current gray wolf individual and the prey position, which is used to characterize the gap between the gray wolf and the prey position, so as to guide the gray wolf to update its position towards the prey position; t is the current iteration number; A and C are coefficient vectors, A = 2a·r 1 - a, C = 2·r 2 , a is the convergence factor, r 1 and r 2 is a vector of random numbers with a modulus between [0, 1];

[0059] The representation formula of the hunting model includes:

[0060]

[0061] Wherein, C 1 , C 2 and C 3 are random vectors; D α , D β and D δ are the distances between the alpha wolf, beta wolf, and delta wolf and other members of the wolf pack respectively, which are used to characterize the distances between task points; X α , X β and X δ are the positions of the alpha wolf, beta wolf, and delta wolf respectively; X is the position of the current wolf;

[0062] Based on the hunting model and the pursuit model, obtain the inspection sequence of the inspection task points and complete the sequence optimization; wherein, the representation formula of the evaluation function includes:

[0063]

[0064] Wherein, f(X) is the evaluation function; X is the position of the current wolf; γ is the weight coefficient; is the path distance of the solution of the first sequence; is the path distance of the current planned sequence solution; is the path distance required for the starting point to reach the last inspection point with a risk coefficient higher than the risk coefficient threshold;

[0065] In response to the gray wolf algorithm falling into local optimum, modify the convergence factor so that the coefficient vector A is greater than 1 or less than - 1. At this time, the representation formula of the convergence factor includes:

[0066]

[0067] where t is the current iteration number; t max is the maximum number of iterations.

[0068] In a second aspect, an embodiment of the present application provides a multi-UAV mission collaborative planning system based on comprehensive cost constraints, including:

[0069] An initialization module, which is used to obtain an image of the target area and determine the information of task points and the information of UAVs in the image;

[0070] A cost acquisition module, which is used to obtain the distance cost from the UAV to the task point and the path cost between the task points of the UAV;

[0071] A fitness function acquisition module, which is used to obtain a fitness function of the comprehensive cost based on the path cost and the distance cost;

[0072] A task allocation module, which is used to perform task allocation on the UAVs based on the genetic algorithm and in combination with the fitness function to obtain an optimal task allocation;

[0073] A sequence optimization module, which is used to perform sequence optimization on the optimal task allocation based on the grey wolf algorithm to complete the multi-UAV mission collaborative planning.

[0074] Advantageous effects: Compared with the prior art, a multi-UAV mission collaborative planning method provided by an embodiment of the present application first performs data initialization in an image of the target area, determines task point information and UAV information in the image; then determines the distance cost and path cost based on the task point information and UAV information; next, obtains a fitness function of the comprehensive cost by combining the path cost and the distance cost; and performs task allocation on the UAVs through the fitness function based on the genetic algorithm to obtain an optimal task allocation result; finally, performs sequence optimization on the optimal task allocation result based on the grey wolf algorithm to complete the collaborative planning of the multi-UAV mission. For the multi-UAV collaborative planning problem, the present application first adopts a hierarchical coupling method, first predicts the path cost value, and then allocates tasks to each UAV through the genetic algorithm, obtains the optimal task allocation by using the fitness function of the comprehensive cost, so that the overall operation cost is minimized. For the tasks that each UAV needs to execute, multi-constraint coupling is adopted, and while using the grey wolf algorithm to optimize the sequence, the path cost is comprehensively considered, and the priority constraint is used to calculate the cost value, so that the cost required for a single UAV to complete the task is minimized, making the multi-UAV system consume less energy and have higher working efficiency. Description of the Drawings

[0075] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0076] Figure 1 It is a flowchart of the steps of the multi-UAV mission collaborative planning method based on comprehensive cost constraint provided by the embodiment of the present application;

[0077] Figure 2 It is a flowchart of the steps of obtaining path cost in the multi-UAV mission collaborative planning method based on comprehensive cost constraint provided by the embodiment of the present application;

[0078] Figure 3 It is a flowchart of the steps of obtaining the best parent node in the multi-UAV mission collaborative planning method based on comprehensive cost constraint provided by the embodiment of the present application;

[0079] Figure 4 It is a flowchart of the steps of obtaining the optimal task allocation in the multi-UAV mission collaborative planning method based on comprehensive cost constraint provided by the embodiment of the present application;

[0080] Figure 5 It is a flowchart of the steps of obtaining the offspring gene sequence by crossing the parental chromosomes in the multi-UAV mission collaborative planning method based on comprehensive cost constraint provided by the embodiment of the present application;

[0081] Figure 6 It is a module connection diagram of the multi-UAV mission collaborative planning system based on comprehensive cost constraint provided by the embodiment of the present application;

[0082] Figure 7 It is the effect diagram of the final plan;

[0083] Figure 8 It is an explanatory diagram of the path optimization strategy of the equal-length method;

[0084] Figure 9 It is an explanatory diagram of the path optimization strategy of the equal-ratio method;

[0085] Reference numerals: 10, initialization module; 20, cost acquisition module; 30, fitness function acquisition module; 40, task allocation module; 50, sequence optimization module. Detailed implementation manners

[0086] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0087] The task allocation problem refers to allocating corresponding task objectives to each robot and determining the execution order among the objectives under the constraints of the environment and task resources, so as to ensure the optimization of the task execution results. Multi-UAV task allocation is an important link to ensure the execution of multi-UAV tasks and is also a prerequisite for the robot to perform path planning. The coupling of the multi-UAV cooperative path planning and task allocation problem is mainly manifested in several aspects such as multi-constraint optimization coupling, interactive coupling, and cooperative planning coupling. Among them, the interactive coupling comprehensively considers the coordination of task requirements, priority factors, etc., so that the UAV system can complete all tasks more efficiently. For its decoupling strategy, currently there are mainly three types: multi-constraint decoupling strategy and modular decoupling strategy. Multi-constraint decoupling is mostly a numerical optimization method, and its constraint conditions include the shortest path, the lowest energy consumption, and motion constraints. The optimal solution of the multi-constraint problem is obtained through algorithms such as GA (Genetic Algorithm) and SCP (Sequential Convex Programming), but the problem complexity has not been reduced and the computational amount is large; modular decoupling first predicts the path, allocates tasks according to the predicted path cost, and uses the path planning algorithm to plan the path for the tasks of each UAV. This decoupling framework realizes the effect of hierarchical decoupling and reduces the difficulty of the multi-UAV cooperative planning problem, but it is difficult to ensure that all constraint conditions are met in actual operation.

[0088] It can be seen that although remarkable achievements have been made in the multi-UAV cooperative planning problem, it is still difficult to cope with complex real environments. How the multi-UAV system effectively coordinates various optimization objectives in complex task scenarios to achieve the optimal overall performance of the system has certain research value.

[0089] In view of this, an embodiment of the present application provides a multi - UAV task collaborative planning method based on comprehensive cost constraints. First, data initialization is performed on the image of the target area, and task point information and UAV information are determined in the image; then, distance cost and path cost are determined based on the task point information and UAV information; next, a fitness function of the comprehensive cost is obtained by combining the path cost and the distance cost; and based on the genetic algorithm, the UAVs are assigned tasks through the fitness function to obtain the optimal task assignment result; finally, the optimal task assignment result is optimized in sequence based on the gray wolf algorithm to complete the collaborative planning of the multi - UAV task. For the multi - UAV collaborative planning problem, the present application first adopts a hierarchical coupling method, first predicts the path cost value, and then assigns tasks to each UAV through the genetic algorithm. The optimal task assignment is obtained by using the fitness function of the comprehensive cost, so that the overall operation cost is minimized. For the tasks that each UAV needs to execute, multi - constraint coupling is adopted. While optimizing the sequence using the gray wolf algorithm, the path cost is comprehensively considered, and the cost value is calculated by combining the priority constraint, so that the cost required for a single UAV to complete the task is minimized, making the multi - UAV system consume less energy and have higher working efficiency.

[0090] In some embodiments, please refer to Figure 1 , Figure 1 which is the step - flow chart of the multi - UAV task collaborative planning method based on comprehensive cost constraints provided by the embodiments of the present application. The multi - UAV task collaborative planning method based on comprehensive cost constraints provided by the embodiments of the present application is specifically implemented through steps 100 to 500:

[0091] Step 100: Obtain an image of the target area, and determine the information of task points and UAVs in the image.

[0092] Step 200: Obtain the distance cost from the UAV to the task point and the path cost between the task points of the UAV.

[0093] In some embodiments, please refer to Figure 2 , Figure 2 which is the step - flow chart of obtaining the path cost in the multi - UAV task collaborative planning method based on comprehensive cost constraints provided by the embodiments of the present application. The method of obtaining the path cost is specifically implemented through steps 210 to 260:

[0094] Step 210: Determine the starting point and the ending point, and establish search trees at both the starting point and the ending point.

[0095] Step 220: Determine the expansion tree and the connection tree in the two search trees.

[0096] Step 230: Randomly select sampling points in the image through the rapidly - exploring random tree connection algorithm as new nodes.

[0097] Step 240: In the expansion tree, trace back each node, and obtain the best parent node of the new node based on the cross - product collision detection algorithm.

[0098] In some embodiments, refer to Figure 3 , Figure 3 which is the flowchart of the step of obtaining the best parent node in the multi - UAV mission collaborative planning method based on comprehensive cost constraints provided by the embodiments of this application. The method of obtaining the best parent node is specifically implemented through steps 241 to 246:

[0099] Step 241: Determine the vector formed by the obstacle edge segments as the obstacle edge segment vector.

[0100] Step 242: Determine the vector formed by the connection line between the new node and the nodes in the expansion tree as the path segment vector.

[0101] Step 243: Obtain the cross - product of the obstacle edge segment vector and the path segment vector.

[0102] Step 244: In response to the cross - product being greater than zero, determine that there is no collision interference between the obstacle edge segment vector and the path segment vector.

[0103] Step 245: In response to the cross - product being equal to zero and the cross - product of the cross vectors being zero, determine that there is no collision interference between the obstacle edge segment vector and the path segment vector.

[0104] Step 246: Determine the node in the expansion tree corresponding to the minimum path segment vector without collision interference as the best parent node of the new node.

[0105] Step 250: Obtain the direct connection path between the node closest to the new node in the connection tree and the expansion tree as the initial path.

[0106] Step 260: Optimize the initial path, obtain the final path, and obtain the path cost based on the final path.

[0107] In some embodiments, path optimization is performed based on the equidistant method of a triangle: Among the nodes of the initial path, obtain the first node, the second node, and the third node; Connect the second node and the first node to form the first side, and connect the second node and the third node to form the second side; Determine the fourth node on the first side and the fifth node on the second side, and the distances from the fourth node and the fifth node to the second node are the same; Connect the fourth node and the fifth node to obtain the third side; In response to the third side having no collision interference with the obstacle, update the fourth node as the first node and update the fifth node as the third node.

[0108] In some embodiments, path optimization is performed based on a triangular proportional method: among the nodes of the initial path, a first node, a second node, and a third node are obtained; a first edge is formed by connecting the second node and the first node, and a second edge is formed by connecting the second node and the third node; a fourth node is determined on the first edge, and a fifth node is determined on the second edge; a third edge is formed by connecting the fourth node and the second node, and a fourth edge is formed by connecting the fifth node and the second node, where the ratio of the length of the third edge to the length of the first edge is a, and the ratio of the length of the fourth edge to the length of the second edge is b, and a = b is satisfied; the fourth node and the fifth node are connected to obtain a fifth edge; in response to the fifth edge having no collision interference with the obstacle, the fourth node is updated to the first node, and the fifth node is updated to the third node.

[0109] Step 300: Based on the path cost and the distance cost, obtain the fitness function of the comprehensive cost.

[0110] Step 400: Based on the genetic algorithm and in combination with the fitness function, perform task allocation for the UAVs to obtain the optimal task allocation.

[0111] In some embodiments, please refer to Figure 4 , Figure 4 is the flowchart of the steps for obtaining the optimal task allocation in the multi-UAV task collaborative planning method based on comprehensive cost constraints provided by the embodiments of the present application. The method for obtaining the optimal task allocation is specifically implemented through steps 410 to 460:

[0112] Step 410: Based on the genetic algorithm, initialize the population information in combination with the data of the UAVs and the data of the task points, and construct a gene sequence based on the population information.

[0113] In some embodiments, the representation formula of the gene sequence includes:

[0114]

[0115] where pop is the gene sequence; k is the number of chromosomes, used to represent the number of preset solutions; n is the total number of tasks; a kn is the task point serial number.

[0116] Step 420: Obtain the fitness of the chromosomes in the gene sequence based on the fitness function of the comprehensive cost.

[0117] Step 430: Select the parent chromosomes in the gene sequence based on the fitness.

[0118] Step 440: Perform crossover based on the parent chromosomes to obtain the offspring gene sequence.

[0119] In some embodiments, please refer to Figure 5 , Figure 5The flowchart shows the steps of obtaining the offspring gene sequence by crossing the parental chromosomes in the multi-UAV mission collaborative planning method based on comprehensive cost constraints provided by the embodiments of this application. The method of obtaining the offspring gene sequence by crossing the parental chromosomes is specifically implemented through steps 441 to 443:

[0120] Step 441: Based on the partially matched crossover method, determine two crossover points in the parental chromosomes.

[0121] Step 442: Determine the crossover region based on the two crossover points.

[0122] Step 443: Exchange the crossover region and determine the mapping relationship to obtain the offspring gene sequence.

[0123] Step 450: Based on the reverse mutation method, perform a mutation operation on the chromosomes in the offspring gene sequence to obtain a new generation of population.

[0124] Step 460: In response to the new generation of population meeting the population threshold, determine the optimal task allocation based on the new generation of population.

[0125] Step 500: Based on the grey wolf algorithm, optimize the sequence of the optimal task allocation to complete the multi-UAV mission collaborative planning.

[0126] In some embodiments, first determine that the preset solution for sequence optimization is the position data of the grey wolf;

[0127] Then determine that the number of preset solutions for sequence optimization is the population size of the grey wolf;

[0128] Then, based on the position data and the population size, determine the hunting model and the pursuit model. Among them, the representation formula of the hunting model includes:

[0129] D = |C · X p (t) - X(t)|;

[0130] X(t + 1) = X p (t) - A · D;

[0131] Where, X p is the position vector of the prey; X is the position vector of the grey wolf, D represents the distance between the current grey wolf individual and the prey position, which is used to characterize the gap between the grey wolf and the prey position, so as to guide the grey wolf to update its position towards the prey position; t is the current iteration number; A and C are coefficient vectors, C = 2a · r 1 - a, C = 2 · r 2 , a is the convergence factor, r 1 and r 2 are vectors of random numbers with a modulus between [0, 1]; X p is the position vector of the prey; X is the position vector of the grey wolf;

[0132] The representation formula of the hunting model includes:

[0133]

[0134] Among them, C 1 , C 2 and C 3 are random vectors; D α , D β and D δ are the distances between the alpha wolf, beta wolf, and delta wolf and the other members of the wolf pack respectively, used to represent the distances between task points; X α , X β and X δ are the positions of the alpha wolf, beta wolf, and delta wolf respectively; X is the position of the current wolf.

[0135] Based on the hunting model and the pursuit model, obtain the inspection sequence of the inspection task points and complete sequence optimization; among them, the representation formula of the evaluation function includes:

[0136]

[0137] Among them, f(X) is the evaluation function; X is the position of the current wolf; γ is the weight coefficient; is the path distance of the first sequence solution; is the path distance of the current planned sequence solution; is the path distance required from the starting point to the last inspection point with a risk coefficient higher than the risk coefficient threshold;

[0138] In response to the gray wolf algorithm falling into a local optimum, modify the convergence factor so that the coefficient vector A is greater than 1 or less than -1. At this time, the representation formula of the convergence factor includes:

[0139]

[0140] Among them, t is the current iteration number; t max is the maximum iteration number.

[0141] Understandably, for the multi-UAV mission collaborative planning method based on comprehensive cost constraints provided by the embodiments of the present application, data initialization is first performed on the image of the target area, and task point information and UAV information are determined in the image; then the distance cost and path cost are determined based on the task point information and UAV information; next, the fitness function of the comprehensive cost is obtained by combining the path cost and the distance cost; and the tasks are assigned to the UAVs based on the fitness function through the genetic algorithm to obtain the optimal task assignment result; finally, the optimal task assignment result is optimized in sequence based on the gray wolf algorithm to complete the collaborative planning of the multi-UAV mission. For the problem of multi-UAV collaborative planning, the present application first adopts a hierarchical coupling method, first predicts the path cost value, and then assigns tasks to each UAV through the genetic algorithm, and obtains the optimal task assignment by using the fitness function of the comprehensive cost, so that the overall operation cost is minimized. For the tasks that each UAV needs to execute, multi-constraint coupling is adopted, and while optimizing the sequence using the gray wolf algorithm, the path cost is comprehensively considered, and the cost value is calculated by priority constraints, so that the cost required for a single UAV to complete the task is minimized, making the multi-UAV system consume less energy and have higher working efficiency.

[0142] Exemplarily, an embodiment of the present application provides a specific implementation scheme of a UAV mission collaborative planning method, and its specific steps include:

[0143] (1) Description and model establishment of the multi-UAV task assignment problem:

[0144] A. Map simulation:

[0145] Simulate a 500*500 two-dimensional view of the real environment according to the real environment. Please refer to Figure 7 , Figure 7 which is the final planned effect diagram, where the obstacles are black squares, the feasible space is white spaces, and 20 position coordinates are randomly set as task points, that is, inspection task points.

[0146] B. UAV information model:

[0147] Assume that there are a total of m UAVs, denoted as V = {V i |i = 1, 2...m}, establish a two-dimensional coordinate system in the two-dimensional view, and the two-dimensional coordinate system includes a vertically arranged x-axis and y-axis. Then the starting position information of each UAV is P i =(x i , y i ), x i is the x-axis coordinate of the starting position of UAV i, and y i is the y-axis coordinate of the starting position of UAV i.

[0148] C. Task point information model:

[0149] Suppose there are n task points, and the relevant information model is represented as P = {x i , y i , d}, i = 1, 2... n, where d is a flag for higher priority. When d = 1, it indicates that the task point has a higher priority, and when d = 0, the priority is lower; x i is the x-axis coordinate of the task point, and y i is the y-axis coordinate of the task point.

[0150] The cost required for the UAV to plan the path between any two task points j 1 , j 2 is E p (j 1 , j 2 ). The cost required between the starting position of any UAV i and any task point j is denoted as E v (i, j).

[0151] D. Genetic algorithm model:

[0152] The genetic algorithm is used for task allocation. Initialize the population information. Define the task length matrix Tasks = {a, b, c, d, e} (a + b + c + d + e = n). Among them, Tasks(i) is the number of tasks executed by the i-th UAV, and the representation formula of its gene sequence is

[0153]

[0154] where op is the gene sequence; k is the number of chromosomes, that is, the number of initial preset solutions; a kn is the task point serial number. Each chromosome in this gene sequence, that is, each row vector, is obtained by arranging the task point serial numbers, and each serial number has and only has one; n is the total number of tasks.

[0155] Establish a fitness matrix to record the fitness of the current k chromosomes, so as to retain the chromosomes with higher fitness and eliminate the chromosomes with lower fitness. The fitness matrix is denoted as adaptability = {f1, f2,..., fk}, where fk is the fitness of the k-th chromosome;

[0156] Calculate the probability of each chromosome being selected in the next evolution according to the fitness matrix, denoted as selectionProbability = {s1, s2,..., sk}, where sk is the probability of the k-th chromosome being selected; its representation formula includes:

[0157] selectionProbability[i] = adaptability[i] / sum(adaptability);

[0158] To endow it with optimization capabilities, a crossover method based on task exchange is combined with a cycle crossover operator based on position crossover. The cycle crossover can regularly recombine individuals through cycling, break the fixed positions of local optimal solutions, and thus jump out of the local optimum. The crossover method based on tasks also makes up for the shortcoming of the cycle crossover's insufficient utilization of task characteristics.

[0159] First, select two parent individuals from the population, that is, select two solutions with higher fitness, randomly select tasks for exchange, and obtain a temporary individual after the exchange. Perform a cycle crossover operation on the temporary individual. First, select a gene on one of the parents. Suppose the selected gene number is x, find the gene number at the same position on the other parent, and then return to the original parent individual to find the position where the gene number is located. Repeat this process until a cycle is formed, and its gene position is the selected position. Retain the genes at the selected positions of parent A' and place the remaining genes of B' into the offspring, thus forming a new offspring.

[0160] (2) Q-RRT and RRT-Connect combined algorithm:

[0161] A. Q-RRT (Quick Randomized Rapid Tree, Q-Quick Random Tree Algorithm) and RRT-Connect (Rapidly-exploring Random Tree Connect, Rapid Random Tree Connection Algorithm) combined algorithm

[0162] First, use the image edge processing function to process the map information. The main steps are to dilate the map. The dilation radius depends on the radius of the moving UAV body, which can not only ensure path safety but also make it easier to obtain the edge information of circular obstacles. Search trees T a ,T b are established at the starting point and the end point respectively. Start iterative sampling. Start RRT-Connect to randomly select sampling points, and determine the new node x according to the position where the sampling point is located new , obtain x new . Then, according to the backtracking idea, check whether a smaller path cost can be generated with an earlier parent node without conflict. If it exists, use this node as the best parent node. If not, proceed to the next step. Check whether there is a node in the search tree other than this node that can be directly connected to this node. If it exists, it means a solution is found. After obtaining the initial path, adopt a path optimization strategy, that is, use the triangle inequality to optimize the initial path with equal proportion and equal distance to reduce the path cost. The algorithm flow is shown in Table 1.

[0163] Table 1. Algorithm flow of the Q-RRT–Connect function

[0164]

[0165]

[0166] B. Cross-Product Based Collision Detection Algorithm:

[0167] The CollisionFree function (Collision-Free function) is used for collision detection. First, define the vector as the vector of the obstacle edge segment, and the vector as the vector of the detection path segment. By calculating their cross product, if the result is greater than 0, it proves that no collision has occurred; when the result is less than 0, it means that the two line segments must intersect, that is, a collision has occurred; when the result is equal to 0, and there is a collinear situation among the three points, it means that one end point of a line segment is on the other line segment, that is, a collision has occurred; if the result is equal to 0 and the cross product of the cross vector is 0, it proves that the two line segments are parallel and no collision occurs, and the path is feasible, otherwise the two line segments are not parallel and do not intersect, and the path is feasible. The CollisionFree function is shown in Table 2.

[0168] Table 2. Algorithm Flow of CollisionFree Function

[0169]

[0170]

[0171] Among them, the DIRECTION function (direction function) is used to calculate the cross product, and the ON-SEGMENT function (function to judge points on the line segment) is used to calculate the intersection of several line segments formed by the edge information of the map obstacles and the sub-path in turn to detect whether there is a collision with the obstacles, that is, whether a collision has occurred.

[0172] C. Path Optimization Strategy:

[0173] Please refer to Figure 8 and Figure 9 , Figure 8 which are the explanatory diagrams of the path optimization strategy of the equal-distance method. Figure 9 which are the explanatory diagrams of the path optimization strategy of the equal-ratio method. For any node in the initial path nodes, randomly select its previous and next nodes to construct a triangle, denoted as: △V i-1 V i V i+1 ; Please refer to Figure 8 , Figure 8 which are the explanatory diagrams of the path optimization strategy of the equal-distance method. In the equal-distance method, define the length e. Starting from V i , determine points V i ' and V i-1 ' on both sides with V i+1 as the vertices at a distance of e, and connect them to form △V i-1 'Vi V i+1 '; Please refer to Figure 9 , Figure 9 , which is an explanatory diagram of the path optimization strategy for the equal-proportion method. In the equal-proportion method, a ratio p (p ∈ (0, 1)) is defined. Taking p as the ratio, V i is used as the vertex to construct △V i-1 ”V i V i+1 ”. Take the points on the edge V i-1 'V i+1 ' and the edge V i-1 ”V i+1 ” to detect whether the nodes collide with obstacles. If there is no collision, define the triangle where the non-colliding edge is located as △V i-1 V i V i+1 , and perform multiple loop iterations, alternately using the equal-distance method and the equal-proportion method to reduce the path cost.

[0174] (3) Task allocation algorithm based on genetic algorithm:

[0175] A. Genetic algorithm:

[0176] Allocate tasks through the genetic algorithm. First, initialize the gene sequence information, randomly define the gene sequence pop and the corresponding task length matrix. Evaluate the individual fitness of each gene sequence according to the fitness function considering the comprehensive cost, denoted as adaptability. Select genes as parent genes in the gene sequence with a probability of selectionProbability. The higher the fitness of the gene sequence, the greater the probability of being selected, while the gene sequences with lower fitness are eliminated; cross the parent genes according to a certain method to produce offspring gene sequences; mutate the offspring chromosomes; generate a new generation of populations, and re-evaluate the fitness until the optimal solution is generated.

[0177] B. Fitness function:

[0178] The fitness function comprehensively calculates information such as the priority level and the path distance cost to obtain the fitness value, which is inversely proportional to the path cost and inversely proportional to the final inspection sequence number of the task points with higher priority. Its formula is

[0179]

[0180] where f is the cost function for calculating the fitness value of the preset solution during the iteration process; j is the current preset solution, which allocates the tasks to m groups, and the value of its cost function is the sum of the costs of each group; dan_n is the sequence position of the task point with a higher danger coefficient in the current nth group; len_n is the total path cost value of the current nth group.

[0181] C. Crossover operator:

[0182] The crossover method adopts the partially matched crossover method. The crossover area is determined by randomly selecting two crossover points. According to the one-to-one mapping relationship between the two sets of genes obtained by swapping, the offspring genes are modified to ensure that there are no conflicts in the new pair of offspring genes.

[0183] D. Mutation operator:

[0184] The mutation method adopts the reverse order mutation method. By randomly selecting a segment of the sequence in an individual and changing the values of this segment of the sequence to the reverse order to generate a new individual, gene duplication is avoided.

[0185] (4) Sequence optimization algorithm based on the grey wolf algorithm:

[0186] A. Grey wolf optimization algorithm

[0187] The Grey Wolf Optimize (GWO) algorithm is a swarm intelligence optimization algorithm formed by simulating the foraging behavior of wolf packs in the biological world. Wolf packs have a strict social hierarchy and organizational system. The top wolf in the wolf pack is the α wolf (the leader of the wolf pack), the β wolf is the subordinate wolf of the α wolf and follows the instructions of the α wolf. The δ wolf has a status second only to the α wolf and the β wolf, and the rest of the wolves are ω wolves with the lowest status. The hunting of wolf packs mainly includes three stages: searching, surrounding, and attacking. In the mathematical model of the grey wolf optimization GWO algorithm, each grey wolf represents 1 candidate solution in the population. The optimal solution is regarded as the α wolf, the second and third best candidate solutions are regarded as the β wolf and the δ wolf, and the remaining candidate solutions are regarded as ω. What is solved in the present invention is the inspection sequence. Assuming that the population size of grey wolves is N (i.e., the number of preset solutions of the algorithm) and the search space is a Dim-dimensional space (i.e., the number of inspection points), the position vector of the i-th grey wolf in the Dim-dimensional space can be expressed as X i Dim , and the mathematical model for grey wolves to surround and capture prey is described as follows:

[0188] D = |C·X p (t) - X(t)|;

[0189] X(t + 1) = X p (t) - A·D;

[0190] Among them, Xp is the position vector of the prey; X is the position vector of the grey wolf. D represents the distance between the current grey wolf individual and the prey position, which is used to characterize the gap between the grey wolf and the prey position to guide the grey wolf to update its position towards the prey position; t represents the current iteration number; A and C are coefficient vectors; that is, the currently preset position information. The calculation formulas for the coefficient vectors A and C are:

[0191] A = 2a·r 1 - a;

[0192] C = 2·r 2 ;

[0193] where: r 1 and r 2 are vectors of random numbers with a modulus between [0, 1]; a is the convergence factor, which linearly decreases from 2 to 0 as the number of iterations increases.

[0194]

[0195] where: t is the current iteration number; t max is the maximum number of iterations.

[0196] The mathematical model for the gray wolves to track the prey's position is described as follows:

[0197]

[0198] where: C 1 , C 2 and C 3 are random vectors; D α , D β and D δ are the distances between the alpha wolf, beta wolf, and delta wolf and the other members of the wolf pack respectively, used to characterize the distances between task points; X α , X β and X δ are the positions of the alpha wolf, beta wolf, and delta wolf respectively; X is the position of the current wolf.

[0199] B. Improved gray wolf optimization algorithm:

[0200] When the prey stops moving, the gray wolves complete the hunting process by attacking. To simulate approaching the prey, the value of the convergence factor a is gradually decreased, so the fluctuation range of the coefficient vector A also decreases accordingly. That is, during the iteration process, when the value of the convergence factor a linearly decreases from 2 to 0, the value of its corresponding coefficient vector A also changes within the interval [-a, a]. When the value of the coefficient vector A is within the interval [-a, a], the next position of the gray wolf can be at any position between its current position and the prey's position. When -1 < A < 1, the wolf pack attacks the prey and falls into a local optimum. When A > 1 or A < -1, the gray wolves move away from the prey and explore other areas to continue searching for the global optimum solution.

[0201] To avoid the algorithm falling into a local optimum, a local optimum detection algorithm is added to the algorithm. The detection algorithm obtains the number of times the same solution is obtained. If the number of iterations to obtain the same solution exceeds a certain limit, it is determined that the algorithm has fallen into a local optimum. After determining that the algorithm has fallen into a local optimum, the value of the convergence factor a is modified according to the current iteration number t, such that A > 1 or A < -1. The calculation formula for the value of a at this time includes:

[0202]

[0203] Meanwhile, to avoid the risk that the sequence solution given by the algorithm does not preferentially inspect the task points with higher risk coefficients during the execution process, the evaluation function of its optimization algorithm is modified in the present invention, and the evaluation function is as follows:

[0204]

[0205] Among them, γ is the weight coefficient. first_length is the path distance of the first sequence solution, length is the path distance of the current planned sequence solution, and danger_length is the path distance required to reach the last inspection point with a higher risk coefficient from the starting point.

[0206] In the experimental map, the number of inspection points is set to 20, the number of inspection points with higher risk coefficients is 5, the population size of the grey wolf algorithm is 100, and the number of genes is 100. The tasks are allocated and planned to prove that the cost required to complete the allocated tasks using this method is the smallest, and while the cost for a single drone to complete the tasks is small, it preferentially inspects the points with higher priorities. The planning results are as Figure 7 shown, where the task points 1, 2, 8, 11, and 16 are the task points with higher risk systems.

[0207] Correspondingly, an embodiment of the present application further provides a UAV task collaborative planning system. Please refer to Figure 6 , Figure 6 which is the module connection diagram of the UAV task collaborative planning system provided by the embodiment of the present application. The UAV task collaborative planning system provided by the embodiment of the present application includes:

[0208] An initialization module 10, which is used to obtain an image of the target area and determine the information of the task points and the information of the UAVs in the image;

[0209] A cost acquisition module 20, which is used to obtain the distance cost of the UAV to the task points and the path cost between the task points;

[0210] A fitness function acquisition module 30, which is used to obtain the fitness function of the comprehensive cost based on the path cost and the distance cost;

[0211] A task allocation module 40, which is used to allocate tasks to the UAVs based on the genetic algorithm and in combination with the fitness function to obtain the optimal task allocation;

[0212] A sequence optimization module 50, which is used to perform sequence optimization on the optimal task allocation based on the grey wolf algorithm to complete the multi-UAV task collaborative planning.

[0213] In some embodiments, the cost acquisition module 20 is specifically configured to:

[0214] Determine a starting point and an ending point, and establish search trees at both the starting point and the ending point;

[0215] Determine an expansion tree and a connection tree in the two search trees;

[0216] Randomly select sampling points in the image through the rapidly-exploring random tree connection algorithm as new nodes;

[0217] In the expansion tree, trace back each node, and obtain the best parent node of the new node based on the cross-product-based collision detection algorithm;

[0218] Obtain the direct connection path between the node closest to the new node in the connection tree and the expansion tree as the initial path;

[0219] Optimize the initial path, obtain the final path, and obtain the path cost based on the final path.

[0220] In some embodiments, the cost acquisition module 20 is specifically configured to:

[0221] Determine the vector formed by the obstacle edge line segments as the obstacle edge line segment vector;

[0222] Determine the vector formed by the connection line between the new node and the nodes in the expansion tree as the path line segment vector;

[0223] Obtain the cross product of the obstacle edge line segment vector and the path line segment vector;

[0224] In response to the cross product being greater than zero, determine that there is no collision interference between the obstacle edge line segment vector and the path line segment vector;

[0225] In response to the cross product being equal to zero and the cross product of the cross vectors being zero, determine that there is no collision interference between the obstacle edge line segment vector and the path line segment vector;

[0226] Determine the node in the expansion tree corresponding to the minimum path line segment vector without collision interference as the best parent node of the new node.

[0227] In some embodiments, the cost acquisition module 20 is specifically configured to:

[0228] Among the nodes of the initial path, obtain a first node, a second node, and a third node;

[0229] Connect the second node and the first node to form a first side, and connect the second node and the third node to form a second side;

[0230] Determine a fourth node on the first side and a fifth node on the second side, and the distances from the fourth node and the fifth node to the second node are the same;

[0231] Connect the fourth node and the fifth node to obtain the third side;

[0232] In response to the third side having no collision interference with the obstacle, update the fourth node to the first node and update the fifth node to the third node.

[0233] In some embodiments, the cost acquisition module 20 is specifically configured to:

[0234] Among the nodes of the initial path, obtain the first node, the second node, and the third node;

[0235] Connect the second node and the first node to form the first side, and connect the second node and the third node to form the second side;

[0236] Determine the fourth node on the first side and the fifth node on the second side;

[0237] Connect the fourth node and the second node to form the third side, connect the fifth node and the second node to form the fourth side, the ratio of the length of the third side to the length of the first side is a, and the ratio of the length of the fourth side to the length of the second side is b, satisfying: a = b;

[0238] Connect the fourth node and the fifth node to obtain the fifth side;

[0239] In response to the fifth side having no collision interference with the obstacle, update the fourth node to the first node and update the fifth node to the third node.

[0240] In some embodiments, the fitness function acquisition module 30 is specifically configured to:

[0241] Based on the genetic algorithm, initialize the population information by combining the data of the unmanned aerial vehicle and the data of the task points, and construct the gene sequence based on the population information;

[0242] Obtain the fitness of the chromosomes in the gene sequence based on the fitness function of the comprehensive cost;

[0243] Select the parent chromosomes in the gene sequence based on the fitness;

[0244] Perform crossover based on the parent chromosomes to obtain the offspring gene sequence;

[0245] Perform mutation operations on the chromosomes in the offspring gene sequence based on the reverse order mutation method to obtain a new generation of population;

[0246] In response to the new generation of population meeting the population threshold, determine the optimal task allocation based on the new generation of population.

[0247] In some embodiments, the fitness function acquisition module 30 is specifically configured to:

[0248] Determine two crossover points in the parental chromosomes based on the partial matching crossover method;

[0249] Determine the crossover region based on the two crossover points;

[0250] Exchange the crossover region and determine the mapping relationship to obtain the offspring gene sequence.

[0251] In some embodiments, the sequence optimization module 50 is specifically configured to:

[0252] Determine that the preset solution for sequence optimization is the position data of the gray wolves;

[0253] Determine that the number of preset solutions for sequence optimization is the population size of the gray wolves;

[0254] Determine the hunting model and the pursuit model based on the position data and the population size. Among them, the representation formula of the hunting model includes:

[0255] D = |C·X p (t) - X(t)|;

[0256] X(t + 1) = X p (t) - A·D;

[0257] Among them, X p is the position vector of the prey; X is the position vector of the gray wolf, D represents the distance between the current gray wolf individual and the prey position, and is used to characterize the gap between the gray wolf and the prey position to guide the gray wolf to update its position towards the prey position; t is the current iteration number; A and C are coefficient vectors, A = 2a·r 1 - a, C = 2·r 2 , a is the convergence factor, r 1 and r 2 are vectors of random numbers whose modulus is between [0, 1]; X p is the position vector of the prey; X is the position vector of the gray wolf;

[0258] The representation formula of the pursuit model includes:

[0259]

[0260] Among them, C 1 , C 2 and C 3 are random vectors; D α , D β and D δ are the distances between the alpha wolf, beta wolf, and delta wolf and the other members of the wolf pack respectively, and are used to characterize the distances between the task points; X α , X β and X δ are the positions of the alpha wolf, beta wolf, and delta wolf respectively; X is the position of the current wolf;

[0261] Based on the hunting model and the pursuit model, obtain the inspection sequence of the inspection task points and complete the sequence optimization; among them, the representation formula of the evaluation function includes:

[0262]

[0263] Among them, f(X) is the evaluation function; X is the position of the current wolf; γ is the weight coefficient; first_length is the path distance of the first sequence solution; lenth is the path distance of the current planned sequence solution; danger_length is the path distance required from the starting point to the last inspection point whose risk coefficient is higher than the risk coefficient threshold;

[0264] In response to the gray wolf algorithm falling into local optimality, modify the convergence factor so that the coefficient vector A is greater than 1 or less than -1. At this time, the representation formula of the convergence factor includes:

[0265]

[0266] Among them, t is the current iteration number; t max is the maximum iteration number.

[0267] The above of this application has introduced in detail a multi-UAV task collaborative planning method and system based on comprehensive cost constraints provided by the embodiments of this application. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those skilled in the art, according to the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A multi-UAV task collaborative planning method based on comprehensive cost constraints, characterized in that: include: Acquire an image of the target area, and determine information of the mission point and information of the drone in the image; Obtaining the distance cost from the drone to the mission point, and the path cost of the drone between the mission points; Based on the path cost and the distance cost, obtaining a fitness function of the comprehensive cost; Based on the genetic algorithm and in combination with the fitness function, the tasks of the UAV are assigned to obtain the optimal task assignment; The optimal task allocation is sequentially optimized based on the Grey Wolf Algorithm to complete the collaborative planning of multi-UAV tasks.

2. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 1 is characterized in that: The steps of obtaining the path cost include: Determine a starting point and an end point, and establish a search tree at the starting point and the end point; Determining an expansion tree and a connection tree in the two search trees; Randomly selecting sampling points in the image by a fast random tree connection algorithm as new nodes; In the extended tree, each node is traced back, and the best parent node of the new node is obtained based on a cross product collision detection algorithm; Obtaining a direct connection path between a node in the connection tree closest to the new node and the expansion tree as an initial path; The initial path is optimized to obtain a final path, and the path cost is obtained based on the final path.

3. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 2 is characterized in that: The step of obtaining the best parent node of the new node based on the cross product collision detection algorithm includes: Determine the vector formed by the obstacle edge line segments as the obstacle edge line segment vector; Determine that a vector formed by a line connecting the new node and nodes in the extended tree is a path segment vector; Obtaining the cross product of the obstacle edge line segment vector and the path line segment vector; In response to the cross product being greater than zero, determining that the obstacle edge line segment vector and the path line segment vector have no collision interference; In response to the cross product being equal to zero and the cross vector cross product being zero, determining that the obstacle edge line segment vector and the path line segment vector have no collision interference; The node in the expanded tree corresponding to the minimum path segment vector without collision interference is determined as the best parent node of the new node.

4. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 2 is characterized in that: The step of optimizing the initial path comprises: Among the nodes of the initial path, obtain a first node, a second node, and a third node; Connecting the second node and the first node to form a first edge, and connecting the second node and the third node to form a second edge; Determine a fourth node on the first side, determine a fifth node on the second side, and the fourth node and the fifth node have the same distance from the second node; Connect the fourth node and the fifth node to obtain a third edge; In response to the third edge having no collision interference with the obstacle, the fourth node is updated to the first node, and the fifth node is updated to the third node.

5. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 2 is characterized in that: The step of optimizing the initial path comprises: Among the nodes of the initial path, obtain a first node, a second node, and a third node; Connecting the second node and the first node to form a first edge, and connecting the second node and the third node to form a second edge; Determine a fourth node on the first edge, and determine a fifth node on the second edge; The fourth node and the second node are connected to form a third side, the fifth node and the second node are connected to form a fourth side, the ratio of the length of the third side to the length of the first side is a, and the ratio of the length of the fourth side to the length of the second side is b, satisfying: a=b; Connect the fourth node and the fifth node to obtain a fifth edge; In response to the fifth edge having no collision interference with the obstacle, the fourth node is updated to the first node, and the fifth node is updated to the third node.

6. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 1 is characterized in that: The step of performing task assignment on the UAV based on the genetic algorithm and in combination with the fitness function to obtain the optimal task assignment includes: Based on the genetic algorithm, the population information is initialized by combining the data of the UAV and the data of the mission point, and the gene sequence is constructed based on the population information; Obtaining the fitness of the chromosome in the gene sequence based on the fitness function of the comprehensive cost; selecting a parent chromosome in the gene sequence based on the fitness; Crossover is performed based on the parental chromosomes to obtain the gene sequence of the offspring; Based on the reverse mutation method, the chromosomes in the offspring gene sequence are mutated to obtain a new generation of population; In response to the new generation population satisfying a population threshold, the optimal task allocation is determined based on the new generation population.

7. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 6 is characterized in that: The characterization formula of the gene sequence includes: Wherein, pop is the gene sequence; k is the number of chromosomes, used to represent the number of preset solutions; n is the total number of tasks; a kn The task point number.

8. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 6 is characterized in that: The step of performing crossover based on the parental chromosomes to obtain the offspring gene sequence comprises: Based on a partial matching crossover method, two crossover points are determined in the parent chromosomes; determining an intersection area based on the two intersection points; The intersection regions are exchanged and the mapping relationship is determined to obtain the offspring gene sequence.

9. The multi-UAV task collaborative planning method based on comprehensive cost constraints according to claim 1 is characterized in that: The step of sequentially optimizing the optimal task allocation based on the grey wolf algorithm comprises: Determining that the preset solution for the sequence optimization is the position data of the gray wolf; Determining the number of preset solutions for the sequence optimization to be the population size of gray wolves; A hunting model and a pursuit model are determined based on the location data and the population size, wherein the characterization formula of the hunting model includes: D=|C·X p (t)-X(t)|; X(t+1)=X p (t)-A·D; Among them, X p is the position vector of the prey; X is the position vector of the gray wolf, D represents the distance between the current gray wolf individual and the prey position, which is used to characterize the gap between the gray wolf and the prey position, so as to guide the gray wolf to update its position to the prey position; t is the current iteration number; A and C are coefficient vectors, A = 2a·r1-a, C = 2·r2, a is the convergence factor, r1 and r2 are vectors of random numbers whose modulus is between [0,1]; The characterization formula of the hunting model includes: Among them, C1, C2 and C3 are random vectors; D α , D β and D δ are the distances between wolf α, wolf β, and wolf δ and other members of the wolf pack, respectively, and are used to represent the distances between task points; X α , X β With X δ are the positions of wolf α, wolf β and wolf δ respectively; X is the position of the current wolf; Based on the hunting model and the hunting model, the inspection sequence of the inspection task points is obtained and the sequence optimization is completed. The characterization formula of the evaluation function includes: Among them, f(X) is the evaluation function; X is the current position of the wolf; γ is the weight coefficient; first_length is the path distance required for the first sequence solution; length is the path distance of the current planned sequence solution; danger_length is the path distance required from the starting point to the last inspection point with a risk factor higher than the risk factor threshold; In response to the gray wolf algorithm falling into a local optimum, the convergence factor is modified so that the coefficient vector A is greater than 1 or less than -1. At this time, the characterization formula of the convergence factor includes: Where t is the current iteration number; t max is the maximum number of iterations.

10. A multi-UAV task collaborative planning system based on comprehensive cost constraints, characterized in that: include: An initialization module (10), the initialization module (10) being used to obtain an image of a target area and determine information of a mission point and information of a drone in the image; A cost acquisition module (20), the cost acquisition module (20) is used to acquire the distance cost from the drone to the mission point, and the path cost of the drone between the mission points; A fitness function acquisition module (30), the fitness function acquisition module (30) being used to acquire a fitness function of a comprehensive cost based on the path cost and the distance cost; A task allocation module (40), the task allocation module (40) is used to allocate tasks to the drone based on a genetic algorithm and in combination with the fitness function to obtain an optimal task allocation; A sequence optimization module (50) is used to perform sequence optimization on the optimal task allocation based on the grey wolf algorithm to complete multi-UAV task collaborative planning.

Images