Multi-agent cooperative detection multi-area coverage task planning method based on improved discrete grey wolf optimization algorithm

Abstract

The invention provides a multi-agent cooperative detection multi-area coverage task planning method based on an improved discrete grey wolf optimization algorithm, and aims to realize comprehensive coverage detection of a plurality of non-intersecting target areas and remarkably reduce the completion time of a cooperative coverage task at the same time. The method comprises the following steps: firstly, comprehensively considering detection height, view field range and image resolution requirements, and taking minimization of a longest coverage path as a target to construct an accurate mathematical model; furthermore, the invention provides a multi-area coverage path generation method based on an improved rotating caliper method, and the method can accurately position an entrance point and an exit point of each target area, and optimizes a transition path between the target areas and a coverage path in the area. Aiming at the complexity of solving a large-scale problem, a discrete grey wolf optimization algorithm fusing heuristic population initialization, an adaptive iteration updating strategy and a variable neighborhood search operation is designed, so that the task planning efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-agent task planning, and in particular to a multi-agent collaborative detection multi-area coverage task planning method based on an improved discrete grey wolf optimization algorithm. Background Art

[0002] The multi-agent collaborative detection and multi-area coverage task planning problem means that under the premise of meeting certain constraints (such as range constraints, collision and obstacle avoidance, etc.), multiple agents collaborate with each other to determine their own motion paths, traverse multiple target environmental areas through sensor-perceived detection methods, and meet the optimization goal of minimizing the coverage completion time as much as possible while ensuring full coverage of the target area.

[0003] The problem of multi-region coverage task planning for multi-agent collaborative detection can be decomposed into two sub-problems: target area allocation and coverage path planning. At present, most of the research focuses on the coverage task planning of a single target area, while the coverage task planning problem of multiple non-intersecting target areas is less considered. The multi-region coverage task planning problem is more widely used in real-world scenarios. For example, in disaster response and environmental monitoring tasks, there are sometimes multiple non-intersecting sub-areas that need to be covered. Not only the coverage path planning within each area needs to be considered, but also the optimization of the access order between multiple sub-areas needs to be considered, which makes the problem more complicated to solve. At present, the more common solution is to adopt a decoupled optimization strategy to decouple task allocation from path planning and approximate it by designing a heuristic solution algorithm. However, on the one hand, the decoupled optimization strategy may make the path planning after task allocation infeasible or inefficient, making it difficult to achieve global optimization; on the other hand, the complexity of the heuristic algorithm in design and its dependence on specific empirical rules have poor scalability.

[0004] In addition, in actual application scenarios, different target areas may have different image resolution detection requirements to meet the precise identification of different targets. The detection flight altitude of the agent directly affects the field of view and image resolution of the sensor. As the detection altitude increases, the field of view of the sensor expands, but the image resolution decreases accordingly. Therefore, for target areas with higher image resolution requirements, it is necessary to lower the detection altitude to obtain higher quality images. For sea areas with lower image resolution requirements, the detection altitude can be appropriately increased to expand the field of view of the sensor and thus shorten the task completion time. This difference in requirements leads to the heterogeneity of the coverage paths in each target area. The present invention takes this heterogeneity into account, that is, while covering the target area, multiple agents need to reasonably plan the detection altitude to ensure the image resolution requirements of the target area. Summary of the invention

[0005] In view of the above problems and shortcomings, the present invention provides a multi-agent collaborative detection multi-area coverage task planning method based on an improved discrete gray wolf optimization algorithm, which can achieve coupled optimization of coverage task allocation and path generation, improve collaborative coverage effects and shorten coverage task completion time.

[0006] To achieve the above purpose, the present invention comprehensively considers key factors such as detection flight altitude, sensor field of view range and image resolution requirements, takes minimizing the longest flight path as the optimization goal, and constructs a multi-region coverage task planning model. On this basis, a multi-region coverage path generation method based on an improved rotating caliper method is proposed, which can accurately determine the entry and exit points of each target area, and optimize the transition path between target areas and the coverage path within the target area. In view of the high computational complexity of large-scale problems, in order to improve the solution efficiency, a discrete gray wolf optimization algorithm that combines heuristic initialization methods, adaptive iterative update strategies and variable neighborhood search operations is designed. This method can effectively balance the task load between multiple agents, optimize the access order between target areas, and ensure the balance of task allocation and the efficiency of path planning.

[0007] The present invention adopts the following technical solutions:

[0008] The multi-agent collaborative detection multi-area coverage task planning method based on the improved discrete grey wolf optimization algorithm includes the following steps:

[0009] Step 1: Initialize the target area map, agent and sensor model based on the given target area map and the parameter configuration information of each agent and its sensor. Consider key factors such as detection height, field of view, and image resolution requirements, and take minimizing the longest coverage path as the optimization goal to build a multi-area coverage task model and fitness function.

[0010] Step 2: Initialize the gray wolf population based on the heuristic method. Each gray wolf corresponds to a task allocation scheme, and ensure that each target area in each task allocation scheme is visited by at least one intelligent agent.

[0011] Assuming that the number of agents is M and the number of target areas is N, the coding length of each gray wolf is M+N. The coding of the gray wolf individual includes two parts: the target area access sequence and the number of target areas allocated to each agent. The target area access sequence includes the target area sequence corresponding to each agent arranged from small to large according to the agent sequence number, and the target area sequence corresponding to each agent is composed of the target area sequence visited by each agent arranged in the order of access.

[0012] Step 3, calculate the fitness value of each gray wolf, and determine the top three gray wolf individuals α, β, and δ in the group, where the fitness values ​​of α, β, and δ decrease in descending order;

[0013] Step 4: Randomly select a gray wolf individual in the population, perform a CROSS operation on the selected individual and the gray wolf individuals α, β, and δ to obtain a new gray wolf individual, retain the gray wolf individual with a higher fitness value, and obtain three gray wolf individuals after the crossover operation;

[0014] Step 5, generate a random number rand1 between [0,1]. When rand1≤s, use the exchange operator, insertion operator and inversion operator to operate the three gray wolf individuals obtained in step 3, obtain a new gray wolf individual each time, compare the fitness value of the new gray wolf individual obtained after each operation with the three gray wolf individuals obtained in step 3, retain the gray wolf individual with a higher fitness value, and obtain three gray wolf individuals after adaptive probability perturbation control; when rand1>s, use the random perturbation operator to operate the three gray wolf individuals obtained in step 3, compare the fitness value of the three new gray wolf individuals obtained after the operation with the three gray wolf individuals obtained in step 3, retain the gray wolf individual with a higher fitness value, and obtain three gray wolf individuals after adaptive probability perturbation control;

[0015] Where s is the convergence factor, which is used to balance the selection probability of various neighborhood search operators;

[0016] Step 6: Add 1 to the number of iterations to determine whether the maximum number of iterations has been reached. If yes, output the best individual among the three gray wolf individuals obtained in the last iteration and go to the next step. Otherwise, go to step 3 and replace the gray wolf individuals α, β, δ with the three gray wolf individuals obtained after this iteration.

[0017] Step 7: Calculate the optimal path of the task allocation solution corresponding to the optimal individual to obtain the final path. Further, in step 2, the specific method of heuristically initializing the population is:

[0018] First, N target region numbers are randomly selected and assigned to corresponding sets; then, based on the centroid coordinates of each region, the distances between target regions are calculated; finally, the unassigned regions are inserted into the end of each set in turn, and the selection criterion is the region with the closest distance to the last region in the set.

[0019] Furthermore, in step 4, the specific method of the CROSS crossover operation is:

[0020] Step 4.1. Generate a random number rand2. If rand2≤1 / 3, then use α as the individual to be crossed. If rand2≥2 / 3, then use β as the individual to be crossed. If 1 / 3<rand2<2 / 3, then use δ as the individual to be crossed.

[0021] Step 4.2, randomly generate the position of the target area to be crossed, exchange the selected gray wolf individual with the encoded target area to be crossed of the individual to be crossed, obtain two new gray wolf individuals, compare the fitness values ​​of α, β, δ with the two new gray wolf individuals, and select the gray wolf individuals with the top three fitness as the three gray wolf individuals after the crossover operation.

[0022] Furthermore, in step 5, the exchange, insertion and reversal operator operations are respectively:

[0023] The definition of the exchange operator is: randomly select two target area positions from the code of the original gray wolf individual, exchange the target areas of these two positions, and form a new gray wolf individual;

[0024] Insertion is defined as: randomly selecting two target area positions from the code of the original gray wolf individual, inserting the target area of ​​the former position after the target area of ​​the latter position, and forming a new gray wolf individual;

[0025] The definition of the inversion operator is: randomly select two target region positions from the code of the original gray wolf individual, and reverse the target region sequence between the two target region positions to form a new gray wolf individual;

[0026] The definition of the disorder perturbation operator is: under the premise of keeping the length of the target area visit sequence corresponding to each agent consistent, a random interval is selected in the target area sequence of each agent. Then, within this selected interval, the target area sequence is randomly reorganized to form a new gray wolf individual.

[0027] Furthermore, in step 5, the calculation formula of s is:

[0028]

[0029] Where iter is the current iteration number, and u is the preset control factor value.

[0030] After adopting the above technical solution, the present invention has the following advantages compared with the prior art:

[0031] The present invention considers a coverage path planning problem in a practical application scenario, the goal of which is to achieve visual coverage detection of a series of non-intersecting polygonal target areas under the premise of meeting the detection image resolution of the target area. In order to effectively solve the problem, the present invention first models and analyzes the airborne sensor, and derives the mathematical relationship expression between the detection flight altitude and the image resolution. Ensure that the intelligent agents perform visual coverage at the maximum detection flight altitude that meets the image resolution requirements. Furthermore, the present invention innovatively proposes a discrete gray wolf optimization algorithm that integrates multiple strategies. This method can comprehensively consider the detection flight altitude, the access sequence of the target area, and the coverage path of each area to achieve coupled optimization. Through simulation experiments in mission scenarios of different scales, it is verified that the method proposed in the present invention can not only effectively solve the large-scale multi-area coverage task planning problem, but also significantly surpass many existing solutions in performance.

[0032] The present invention is described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is an overall flow chart of the method of the present invention;

[0034] Figure 2 This is a schematic diagram of a multi-agent multi-area visual coverage scenario;

[0035] Figure 3 It is a schematic diagram of the projected area of ​​the photoelectric load;

[0036] Figure 4 This is a schematic diagram of the resolution of the photoelectric payload image;

[0037] Figure 5 This is a schematic diagram of the optimal coverage path for a single area;

[0038] Figure 6 Schematic diagram for coding individual gray wolves;

[0039] Figure 7 It is a schematic diagram of a set-based coding representation method;

[0040] Figure 8 This is a schematic diagram of CROSS cross operation;

[0041] Fig. 9 Schematic diagram of the disorder perturbation operator;

[0042] Fig.10 Schematic diagram of the exchange operator;

[0043] Fig.11 Schematic diagram of the insertion operator;

[0044] Fig.12 Schematic diagram of the inversion operator;

[0045] Fig.13 It is the top view and stereoscopic view of the path planning based on IDGWO under the condition of scenario 1;

[0046] Fig.14 The top view and stereogram of the IDGWO path planning under scenario 2;

[0047] Fig.15 The top view and stereogram of the IDGWO path planning under scenario three;

[0048] Fig.16 This is the box plot of the Monte Carlo simulation results for scenario 1;

[0049] Fig.17 This is the box plot of the Monte Carlo simulation results for scenario 2;

[0050] Fig.18 Box plot of Monte Carlo simulation results for scenario three. DETAILED DESCRIPTION

[0051] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0052] 1 Problem Description and Model Construction

[0053] 1.1 Problem Description

[0054] like Figure 2 As shown, consider the following task scenario: deployment An intelligent agent with a photoelectric sensor The target areas are visually covered, and these areas are all located in the same geographic space and do not overlap with each other. i , i∈[N] represents the i-th target area. The location, size and shape of the area are known. Among them, P i The j-th row of stores the position information of the j-th vertex in region i. Represents the number of vertices in region i. All agents start from the same base and return to the base after completing the task. The location of the base is known and is represented by Sequence number 0 indicates the starting point.

[0055] In practical applications, monitoring tasks in different areas have different requirements for image clarity. Therefore, when the agent moves from one target area to another to perform a task, it should adjust the detection flight altitude according to the new task requirements to ensure that the image resolution requirements of each area are met.

[0056] 1.2 Model Assumptions

[0057] To simplify the model, the following assumptions are made:

[0058] 1) An existing boundary processing method can simplify any irregular concave polygonal region into a convex polygonal region. Therefore, without loss of generality, it is assumed in the present invention that all target regions are convex polygonal regions.

[0059] 2) Assume that each target area is a two-dimensional plane without terrain undulations to avoid the complexity caused by height changes or occlusions.

[0060] 3) It is assumed that within a single target area, the image resolution requirement is uniform and the height and speed of the agent remain constant within the area.

[0061] 4) When the agent transfers between target areas, the flight altitude can be dynamically adjusted according to the image resolution requirements of the target area. This process is not included in the path planning process.

[0062] 5) Assume that the visual coverage task of each target area is completed independently by a single agent. The maximum range of each agent under the maximum energy constraint is L max .

[0063] Based on the above assumptions, the goal of multi-agent collaborative detection and multi-area coverage path planning is to ensure that all target areas can be effectively covered and monitored while meeting the image resolution requirements of each target area and the range limitations of the agents.

[0064] 1.3 Sensor Model

[0065] like Figure 3 As shown in the figure, the projection area of ​​the photoelectric sensor on the sea can be approximated as a rectangle. Under a certain field of view angle, its area is related to the spatial state parameters of the sensor.

[0066] Assume the projected area is S o =W×L. The straight-line distance between the camera and the target area is l. Considering W≈l·γ, Then S o It can be expressed as:

[0067]

[0068] Where h is the detection flight altitude, θ is the visual axis depression angle (the angle between the photoelectric load and the horizontal plane), is the camera lens field of view.

[0069] From the formula, we can see that the projected area S o It is proportional to the square of the detected flight altitude and inversely proportional to the cube of the sine function of the camera's visual axis depression angle. Determined by the camera focal length, considering that the image distortion is low and the stability is good when shooting with a fixed focal length, it is assumed that a fixed focal length is used for visual coverage. Therefore, the projection area is only related to the detection flight altitude and the visual axis depression angle.

[0070] For optical images, the more pixels there are per unit area, the higher the image resolution. Figure 4 As shown in Figure 1, the agent visually covers the target area at height h. The pixel size of the CCD camera is N H ×N V (Unit: pixel / cm). Assume that the projection size of the sea surface target in the horizontal and vertical directions is W0×L0.

[0071] According to the imaging principle, the pixel size corresponding to the target on the image sensor can be expressed as:

[0072]

[0073] Combining the above formula with formula (1), we can get:

[0074]

[0075] Therefore, the number of pixels per unit area corresponding to the target is:

[0076]

[0077] From the formula, we can know that when the photoelectric load pixel size N H ×N V When a certain time is reached, the target resolution of the optoelectronic payload is proportional to the cube of the sine function of the optoelectronic payload's visual axis depression angle, and is proportional to the square of the detection flight altitude and the field of view angle. Inversely proportional. Assume that the image resolution required by the task is not less than R d ,Right now:

[0078] N″ H ×N″ V ≥R d (5)

[0079] Combining formula (4) with formula (5), we can get the maximum detection flight altitude:

[0080]

[0081] To ensure the minimum image distortion during visual coverage, it is assumed that the visual axis depression angle of the photoelectric sensor is always 90°. After determining the lens field of view angle and pixel size parameters, the image resolution requirements of the target area are met by adjusting the flight altitude. In the present invention, each agent performs visual coverage at the maximum detection flight altitude that meets the image resolution requirements. This strategy can not only ensure image clarity but also effectively increase the field of view range and improve the efficiency of task completion.

[0082] 1.4 Fitness function construction

[0083] When constructing the fitness function, five decision variables need to be considered: the target area detection flight altitude, the number of target areas assigned, the target area visit order, the target area entry and exit points, and the target area internal coverage path. These decision variables are coupled with each other and jointly affect the final multi-agent collaborative detection plan. Therefore, the binary decision variable x is introduced ij ,i,j∈[N]∪{0}, when the variable value is 1, it means that the agent is from area S i Move to area S j , otherwise the variable value is 0. 0i Indicates that the agent moves from the base to the target area S i , x i0 Indicates that the agent is located in the target area S i Return to the base. Therefore, the following constraints can be obtained:

[0084]

[0085] Specifically, formulas (7)-(8) ensure that each area is covered only once. Formulas (9)-(10) ensure that the M agents all start from the same base and return to the base after completing the task. Formula (11) is used to prevent the generation of subloops, where |S| represents the total cardinality of the target area set. Formula (12) represents the variable x ij The value that can be assigned.

[0086] In addition, in order to describe the coverage path of each target area, the variable Indicates the target area S i The intelligent agent flies along this path to achieve complete coverage of the target area. im represents the mth path point in this path, n i Indicates the total number of path points on this path, C i Indicates the target area S i The set of all possible covering paths. i1 and Respectively represent the target area S iIn and out points for overwriting. The specified target area is visited in order, by connecting the coverage paths and base S0, we can get a set of trajectories T of M agents covering all target areas, especially |T| = M. Let τ∈T represent the trajectory of one of the agents, S τ represents the set of all target areas covered by the trajectory τ, satisfying the following constraints:

[0087]

[0088] In the formula, Represents the set S τ The target area access sequence, S τ The absolute value of indicates that each agent is assigned at least one target area and at most M target areas. max , then the length of each flight trajectory τ needs to satisfy the following constraints:

[0089]

[0090] Where, L τ represents the total length of the flight trajectory τ, which can be calculated by the following formula:

[0091]

[0092] The formula consists of three parts. The first part represents the distance from the base position of each agent to the entry point of the first target area. The second part represents the path distance of each agent's transition between target areas. The third part represents the coverage path distance of each agent's target area. In the formula, the function D(p1,p2) represents the Euclidean distance from position p1 to position p2, and the function The coverage path c is calculated i The total length of

[0093] Under the condition that the motion parameters of each agent are the same, the completion time of the multi-region collaborative detection task depends on the longest coverage path. Therefore, the present invention takes minimizing the longest coverage path as the optimization goal. The advantage of this strategy is that it can balance the task load among multiple agents and avoid the waiting time caused by the long coverage path of individual agents, thereby improving the overall task execution efficiency. Therefore, the fitness function is constructed as follows:

[0094]

[0095] stConstraint(7)-(12)

[0096] Aiming at the problem of multi-agent collaborative detection and multi-region coverage task planning, this paper proposes a coupled optimization solution strategy, which encodes the target area allocation scheme and the target area access order as the solution space, and gradually approaches the optimal solution by searching for candidate solutions in the search space through a meta-heuristic algorithm. The multi-region coverage path is generated based on the improved rotating caliper method, and the reciprocal of the longest coverage path is used as the fitness value to evaluate the quality of the candidate solution.

[0097] 2 Multi-area coverage path generation method based on improved rotating caliper method

[0098] Vasquez et al. proposed a rotating caliper method to solve the full coverage path planning problem of a single convex area. The basic idea is to traverse the anti-podal point pairs of the convex polygon, find the parallel support line with the smallest distance between its anti-podal points, and cover the target area with parallel paths in the direction of the parallel support line, such as Figure 5 As shown. Since the rotating caliper method is only applicable to generating a coverage path for a single target area, the present invention proposes a method for generating a multi-area coverage path based on the rotating caliper method.

[0099] Assume that for the i-th agent, there are N i The target areas to be covered are obtained by heuristic initialization method. i When the agent reaches the first target area, it starts from the task starting point P. start Start moving to the entry point P1 of the first area to be covered in , follow the parallel coverage mode to scan and traverse the target area, and finally move to the exit point P1 of the first area to be covered out The parallel coverage model can locate a series of waypoints at the edge of the polygon. The combination of the mission starting point and the set of waypoints is the coverage path for the current target area. Based on the rotating caliper method, the shortest full coverage path for the current target area can be obtained, which is expressed as After completing the coverage of a target area, according to the target area access order, based on the greedy strategy, the first exit point P1 of the area to be covered is out Move to the entry point P2 of the second area to be covered in , and so on, and finally return to the task end point P after covering all target areas end By connecting the shortest full coverage paths of each target area, the overall optimal coverage path is obtained.

[0100] 3 Multi-strategy fusion discrete grey wolf optimization algorithm

[0101] 3.1 Standard Grey Wolf Optimization Algorithm

[0102] The standard Grey Wolf Optimization (GWO) algorithm is a swarm intelligence optimization method developed by simulating the social hierarchy and hunting behavior of grey wolves. It has the characteristics of strong convergence, few parameters, and easy implementation. The GWO optimization process mainly includes the social hierarchy stratification, encirclement, hunting, and attacking of prey by grey wolves. Its mathematical model is described as follows:

[0103] (1) Social stratification

[0104] When designing GWO, we first need to build a gray wolf social hierarchy model. Calculate the fitness value of each individual in the initial population, mark the three gray wolves with the best fitness in the population as α, β, δ, and the remaining gray wolves as ω. In other words, the social hierarchy in the gray wolf group is α, β, δ, ω from high to low. The optimization process of GWO is mainly guided by the three best solutions α, β, δ in each generation of the population.

[0105] (2) Encirclement

[0106] The mathematical model of the gray wolf surrounding its prey is:

[0107] X(t+1)C=X P (t)-A⊙D (17)

[0108] D=|C·X P (t)-X(t)| (18)

[0109] Where: t is the current iteration number; ⊙ is the Hadamard product operation; A and C are the synergy coefficient vectors; D is the distance between the prey and the gray wolf. P is the position vector of the prey; X(t) is the position vector of the wolf. The calculation formulas of A and C are:

[0110] A=2·a·r1-a (19)

[0111] C=2·r2 (20)

[0112] Where r1 and r2 are both random numbers in [0,1]; a is the linear convergence factor, which decays linearly from 2 to 0 as the number of iterations increases.

[0113] (3) Hunting

[0114] After completing the encirclement of the prey, the individuals in the wolf pack hunt by updating their own positions. The mathematical model of this process is:

[0115]

[0116] In the formula, X(t+1) represents the updated position of the individual wolf pack, and the forward length and direction of the individual wolf pack are jointly determined by X1(t), X2(t), and X3(t); α (t), X β (t), X δ (t) represents the position vector of wolf α, β, and δ respectively; D α , D β , D δ Respectively represent the distance between α, β, δ and ω wolves, and the calculation formula is as follows:

[0117]

[0118] (4) Attack

[0119] When the prey stops moving, the hunting activity ends and the wolves attack immediately. In the process of building the attack model, according to formula (19), the decrease in the value of a will cause the value of A to fluctuate, that is, A is a random vector in the interval [-a, a]. Where a decreases linearly during the iteration process. When |A| < 1, the prey is attacked.

[0120] 3.2 Improved discrete grey wolf optimization algorithm

[0121] The standard grey wolf optimization method is suitable for solving continuous optimization problems, while the MTSP-HVCPP problem belongs to a combinatorial optimization problem and cannot be directly solved by the GWO method. Therefore, based on the standard grey wolf optimization method, the present invention integrates multiple strategies such as heuristic population initialization, multi-population coevolution, and domain disturbance operation to propose an improved discrete grey wolf optimization algorithm (Improved Discrete Grey Wolf Optimization, IDGWO).

[0122] 3.2.1 Gray wolf individual coding representation

[0123] The multi-agent target area allocation scheme and the target area visit sequence are used as the solution space to encode the gray wolf individuals. Each gray wolf individual in the gray wolf population corresponds to a feasible solution. The present invention adopts a two-part sequence encoding method, such as Figure 6 As shown in the figure. Assuming that the number of agents is M and the number of target areas is N, the encoding length of the gray wolf individual is M+N. The gray wolf individual is divided into two parts by a vertical line. The numbers on the left represent the visit sequence of the N target areas (excluding the starting position), and the numbers on the right represent the number of target areas assigned to each agent.

[0124] After encoding the gray wolf individuals, the optimal gray wolf individuals need to be converted into optimization variables through decoding to obtain the extreme value of the objective function. The decoding method is to allocate the corresponding number of target areas to each agent for visual coverage detection in the order of the target areas on the left side of the vertical line.

[0125] 3.2.2 Heuristic Population Initialization Strategy

[0126] When initializing the population, we use a set-based encoding representation method, such as Figure 6 As shown in the figure, first, M target area sets are generated according to the number of agents; second, the N target areas are allocated to these M sets using a heuristic initialization method according to the optimization goal; finally, the final gray wolf individuals are generated according to the number of target areas in each set.

[0127] The heuristic initialization method first randomly selects M target region numbers and assigns them to the corresponding sets. Then, based on the centroid coordinates of each region, the distance between the target regions is calculated. Unassigned regions are inserted at the end of each set in turn, and the selection criterion is the region closest to the last region in the set. During each iteration, the updated set is checked to reduce the risk of generating infeasible paths. This process continues until all regions have been assigned. Once the region assignment work in each set is completed, the number of target regions that each agent needs to cover is determined based on the number of regions in each set and encoded accordingly.

[0128] 3.2.3 Cross-update and population elimination strategy

[0129] Since the multi-agent multi-region coverage task planning problem is a combinatorial optimization problem, the standard gray wolf optimization algorithm cannot be directly applied to solve it. Therefore, the crossover mutation idea in the genetic method is used to complete the update of the gray wolf position. The formula is as follows:

[0130]

[0131] For each gray wolf individual position update, first select gray wolf individuals α, β, and δ according to the above social hierarchy mechanism, then generate a random number rand between [0, 1], and perform a CROSS operation on the current gray wolf individual position with the position of one of the gray wolf individuals in α, β, and δ according to different values ​​of rand, as follows: Figure 8As shown. Assuming that the number of target areas is 10, numbered 1 to 10, the number of agents is 3, and all agents start from the same initial position, define the gray wolf individuals X1 and X2. Assuming that the crossover position is 3 to 6, select the crossover segment [8,1,4,5] from X1 and replace it in front of X2, then delete the repeated elements in X2 from beginning to end (the last 3 elements are not included), and finally form a new gray wolf individual X2 = [8,1,4,5,7,3,2,6,10,9,2,4,4].

[0132] After each iteration, G gray wolf individuals with the worst fitness are selected for elimination, G = k·population, k is the ratio factor of population elimination, and population is the total number of individuals in the population. New gray wolf individuals are generated through heuristic methods to replace the eliminated gray wolf individuals, and a greedy strategy is adopted to ensure that the fitness value of the newly generated gray wolf individuals is better than the original fitness value. The population elimination strategy can eliminate individuals with poor fitness in the population, avoid the location area fixation of the solution in the later stage of the method, and cause search stagnation, which is conducive to promoting the co-evolution of the population.

[0133] 3.2.4 Variable Neighborhood Search Strategy

[0134] In order to avoid the method from falling into the local optimum, the α, β, and δ individuals of the gray wolf population after the updated position are searched in the neighborhood to obtain better gray wolf individuals, thereby driving the population to update in a better direction. Since the order of visiting the target area has a great influence on the fitness value, a disordered perturbation operator is introduced, and the three classic neighborhood operators of exchange, insertion, and inversion are combined to perform local search. At the same time, in order to coordinate the use of inter-group communication strategy and local perturbation and effectively balance the global and local search of the method, an adaptive probabilistic perturbation control strategy is introduced to select the perturbation mode by probability, so that the perturbation amplitude is large in the early stage of the method, the population diversity is maintained, and the method is prevented from falling into the local optimum; in the middle and late stages of the method, it is inclined to local search to enhance the development ability of the method and improve the convergence speed of the method.

[0135] (1) Random perturbation operator

[0136] like Fig. 9 As shown, for the gray wolf individual X α , first convert it into an encoding form X based on the target region set L According to X L The length of each target area set in , generates the corresponding non-repeating random integer sequence X S , using X S Adjust X L The gene position on the target region is obtained, and finally it is encoded as a new gray wolf individual X′. α .

[0137] (2) Exchange operator

[0138] like Fig.10 As shown, two positions are randomly selected from the gray wolf individual X, and the target areas of these two positions are exchanged to form a new individual X′.

[0139] (3) Insert Operator

[0140] like Fig.11 As shown, two positions are randomly selected from the gray wolf individual X, and the target area of ​​the previous position is inserted after the target area of ​​the latter position to form a new gray wolf individual X′.

[0141] (4) Inversion operator

[0142] like Fig.12 As shown, two positions are randomly selected from the gray wolf individual X, and the target region sequence between the two positions is arranged in reverse order to form a new gray wolf individual X′.

[0143] Obviously, the random perturbation operator is more suitable for fine-tuning search, while the exchange, insertion and reversal operators show their advantages in achieving efficient perturbations with large changes. In order to better balance the exploration and development of the method, the neighborhood operators are adaptively selected and used according to their different perturbation degrees to the solution: in the early stage of the method, more efficient perturbations of larger magnitude are adopted to accelerate convergence; for the later stage with more high-quality individuals, fine-tuning precise search is mainly carried out to improve the solution accuracy. Therefore, the above neighborhood search operators are selected according to probability to generate new solutions. Each time a random number between [0,1] is generated, when rand≤s, exchange, insertion and reversal are used on the gray wolf individuals in turn. When rand>s, the random perturbation operator is used on the gray wolf individuals to generate new solutions. Let s be the convergence factor. In order to balance the selection probability of multiple neighborhood search operators, it is calculated by the following formula:

[0144]

[0145] Where iter is the current iteration number, u is the control factor. Changing the size of u can control the attenuation amplitude of s. The larger u is, the slower s decays. The corresponding method mainly uses transformation, insertion and reversal operations in the early stage, and mainly performs disorder adjustment within the set in the later stage.

[0146] 3.3 Method solution process

[0147] Assume that the size of the gray wolf population is population, the maximum number of iterations of the method operation is max_iter, the number of iterations in which the fitness value of the best individual in the population remains unchanged is stop_iter, the number of iterations of the neighborhood search operation is search_iter, the population elimination ratio factor is k, and the termination condition of the method is that the fitness value of the best gray wolf individual α remains unchanged for stop_iter generations or reaches the maximum number of iterations max_iter. The main steps are as follows:

[0148] Step 1: Initialize method hyperparameters: max_iter, stop_iter, search_iter, population, k, and initialize the gray wolf population.

[0149] Step 2: Calculate the fitness values ​​of all individuals in the population and determine the top three gray wolf individuals α, β, and δ in the group.

[0150] Step 3: Use the roulette method to select gray wolf individuals in the population and perform a CROSS crossover operation on the selected individuals.

[0151] Step 4: Neighborhood search based on adaptive probability perturbation control strategy

[0152] Step 5: If the method reaches the termination condition, output the global optimal solution; otherwise, repeat Step 2 to Step 4.

[0153] 4 Simulation results and analysis

[0154] 4.1 Parameter settings

[0155] The simulation experiment was conducted based on MATLAB R2023b, and the hardware environment was configured with Intel Core i9-11900H, 2.50GHz CPU and 16GB RAM. The parameters in the simulation experiment were set as follows: the photoelectric sensor visual axis depression angle θ = 90°, the lens field of view angle The initial starting point coordinates of the multi-agent are (14km, 18km), the speed is 20m / s, and the maximum range is 40km. Considering the differences in image resolution requirements of the target area, different detection flight altitudes h are set for different target areas. i ,make Three coverage mission scenarios of different scales are set up, a certain number of target areas are randomly generated in the mission domains of different sizes, and different numbers of agents are deployed for collaborative detection. The parameters of each scenario are shown in Table 1.

[0156] Table 1 Multi-agent multi-area coverage scenario parameters

[0157]

[0158] 4.2 Method Comparison

[0159] The present invention adopts the IDGWO method to solve the problem and generates a multi-area coverage path in Figures 13 to 15 In order to verify the superiority of the IDGWO method proposed in this paper, three benchmark methods are selected for comparative analysis. The benchmark methods include two multi-region coverage task planning methods based on improved genetic algorithms: GASC and GA2PC, and a multi-region coverage path planning method that combines heuristic methods and ant colony optimization algorithms: BETR-VAC. These methods have been proven to be effective in solving multi-agent multi-region coverage task planning problems. Table 2 lists the parameter settings of these methods in detail.

[0160] Table 2 Comparison method parameter settings

[0161]

[0162] In order to eliminate the instability of a single experiment, Monte Carlo simulation was used to conduct 100 experiments for each test method under scenario 4. In each experiment, these methods were iterated up to 200 times. In order to compare the results of each method more clearly, the changes in the optimal objective function value were plotted as a box plot, as shown in Figure 16-18 As shown. It can be found that in terms of indicator stability, the method proposed in the present invention is significantly better than other comparative methods, especially when facing large-scale problems, IDGWO still has good convergence. The performance of the four methods is statistically analyzed, and the results are shown in Table 3. According to the results in Table 3, under the conditions of scenario 1, when facing small-scale problems, IDGWO and the three baseline methods can obtain approximate optimal solutions, but the stability of the IDGWO method is significantly better than the other three methods. In addition, the quality of the solution of the GA2PC method is significantly inferior to that of the other methods, indicating that this method is prone to fall into local optimality. As the scale of the problem increases, the superiority of the IDGWO method gradually emerges, and it is superior to the other methods in terms of solution quality and solution efficiency.

[0163] Table 3 Statistical analysis of Monte Carlo simulation results

[0164]

[0165] 5 Conclusion

[0166] The present invention considers a coverage path planning problem in a practical application scenario, the goal of which is to achieve visual coverage detection of a series of non-intersecting polygonal target areas under the premise of meeting the detection image resolution of the target area. In order to effectively solve the problem, the present invention first models and analyzes the airborne sensor, and derives the mathematical relationship expression between the detection flight altitude and the image resolution. Ensure that the intelligent agents perform visual coverage at the maximum detection flight altitude that meets the image resolution requirements. Furthermore, the present invention innovatively proposes a discrete gray wolf optimization algorithm that integrates multiple strategies. This method can comprehensively consider the detection flight altitude, the access sequence of the target area, and the coverage path of each area to achieve coupled optimization. Through simulation experiments in mission scenarios of different scales, it is verified that the method proposed in the present invention can not only effectively solve the large-scale multi-area coverage task planning problem, but also significantly surpass many existing solutions in performance.

[0167] The above is an example of the best implementation of the present invention, and the parts not described in detail are common knowledge of ordinary technicians in the field. The protection scope of the present invention shall be based on the content of the claims, and any equivalent transformation based on the technical enlightenment of the present invention is also within the protection scope of the present invention.

Claims

1. A multi-agent collaborative detection multi-area coverage task planning method based on an improved discrete grey wolf optimization algorithm, characterized in that: The following steps are involved: Step 1: Initialize the target area map, the agent and its sensor model according to the given target area map and the parameter configuration information of each agent and its sensor; construct a multi-area coverage task model and fitness function with the optimization goal of minimizing the longest coverage path; Step 2: Initialize the gray wolf population based on the heuristic method. Each gray wolf corresponds to a task allocation plan, and ensure that each target area in each task allocation plan is covered by at least one intelligent agent. Assuming that the number of agents is M and the number of target areas is N, the coding length of each gray wolf is M+N. The coding of the gray wolf individual includes two parts: the target area access sequence and the number of target areas allocated to each agent. The target area access sequence includes the target area sequence corresponding to each agent arranged from small to large according to the agent sequence number, and the target area sequence corresponding to each agent is composed of the target area sequence numbers visited by each agent arranged in the order of access. Step 3, calculate the fitness value of each gray wolf, and determine the top three gray wolf individuals α, β, and δ in the group, where the fitness values ​​of α, β, and δ decrease in descending order; Step 4: Randomly select a gray wolf individual in the population, perform a CROSS operation on the selected individual and the gray wolf individuals α, β, and δ to obtain a new gray wolf individual, retain the gray wolf individual with a higher fitness value, and obtain three gray wolf individuals after the crossover operation; Step 5, generate a random number rand1 between [0,1]. When rand1≤s, use the exchange operator, insertion operator and inversion operator to operate the three gray wolf individuals obtained in step 4, obtain a new gray wolf individual each time, compare the fitness value of the new gray wolf individual obtained after each operation with the three gray wolf individuals obtained in step 4, retain the gray wolf individual with a higher fitness value, and obtain three gray wolf individuals after adaptive probability perturbation control; when rand1>s, use the random perturbation operator to operate the three gray wolf individuals obtained in step 4, compare the fitness value of the three new gray wolf individuals obtained after the operation with the three gray wolf individuals obtained in step 4, retain the gray wolf individual with a higher fitness value, and obtain three gray wolf individuals after adaptive probability perturbation control; Where s is the convergence factor, which is used to balance the selection probability of various neighborhood search operators; Step 6: Add 1 to the number of iterations to determine whether the maximum number of iterations has been reached. If yes, output the best individual among the three gray wolf individuals obtained in the last iteration and go to the next step. Otherwise, go to step 3 and replace the gray wolf individuals α, β, δ with the three gray wolf individuals obtained after this iteration. Step 7: Calculate the task allocation plan corresponding to the optimal gray wolf individual, generate the optimal coverage path based on the plan based on the improved rotating caliper method, and obtain the final full coverage path.

2. The multi-agent collaborative detection multi-area coverage task planning method based on the improved discrete grey wolf optimization algorithm according to claim 1 is characterized in that: In step 2, the heuristic initialization method is: first, randomly select N target area numbers and assign them to corresponding sets; then, based on the centroid coordinates of each area, calculate the distance between the target areas; finally, the unassigned areas will be inserted into the end of each set in turn, and the selection criterion is the area closest to the last area in the set.

3. The multi-agent collaborative detection multi-area coverage task planning method based on the improved discrete grey wolf optimization algorithm according to claim 1 is characterized in that: In step 4, the specific method of the CROSS crossover operation is: Step 4.1, generate a random number rand2, if rand2≤1 / 3, then use α as the individual to be crossed, if rand2≥2 / 3, then use β as the individual to be crossed, if 1 / 3<rand2<2 / 3, then use δ as the individual to be crossed; Step 4.2, randomly generate the position of the target area to be crossed, exchange the selected gray wolf individual with the encoded target area to be crossed of the individual to be crossed, obtain two new gray wolf individuals, compare the fitness values ​​of α, β, δ with the two new gray wolf individuals, and select the gray wolf individuals with the top three fitness values ​​as the three gray wolf individuals after the crossover operation.

4. The multi-agent collaborative detection multi-area coverage task planning method based on the improved discrete grey wolf optimization algorithm according to claim 1 is characterized in that: In step 5, the exchange, insertion and reversal operator operations are respectively: The definition of the exchange operator is: randomly select two target area positions from the code of the original gray wolf individual, exchange the target areas of these two positions, and form a new gray wolf individual; Insertion is defined as: randomly selecting two target area positions from the code of the original gray wolf individual, inserting the target area of ​​the former position after the target area of ​​the latter position, to form a new gray wolf individual; The definition of the inversion operator is: randomly select two target region positions from the code of the original gray wolf individual, and reverse the target region sequence between the two target region positions to form a new gray wolf individual; The definition of the disorder perturbation operator is: select the target area access sequence corresponding to each intelligent agent from the code of the original gray wolf individual, randomly select an interval while ensuring that the total length of the target area sequence of each intelligent agent remains unchanged, and randomly reorganize the target area sequence within the selected interval to form a new gray wolf individual.

5. The multi-agent collaborative detection multi-area coverage task planning method based on the improved discrete grey wolf optimization algorithm according to claim 1 is characterized in that: The optimal coverage path is generated by using a multi-region coverage path generation method based on an improved rotating caliper method, and the reciprocal of the total length of the path is used as the fitness value of the gray wolf individual. The multi-region coverage path generation method based on the improved rotating caliper method includes the following steps: Step 1.1: Determine the order in which each agent visits different target areas according to the gray wolf's code. Assume that for the i-th agent, there are N i target areas to be covered, and plans have been made for these N i The order of visiting the regions is as follows: When the agent covers the first target region, it starts from the task starting point P. start Start moving to the entry point P1 of the first area to be covered in , follow the reciprocating path to scan and cover the target area, and finally move to the exit point P1 of the first area to be covered out ;Use parallel round-trip coverage mode to fully cover the target area, and generate a series of waypoints by calculating the intersection of the coverage path and the boundary of the target area; Step 1.2: The set of waypoints corresponding to the mission starting point and each target area is the coverage path for the current target area. The shortest path length is used as the optimization goal. Based on the rotating caliper method, the optimal coverage path for the current area is obtained, which is expressed as After completing the coverage of a target area, according to the target area access order, based on the greedy strategy, the first exit point P1 of the area to be covered is out Move to the next entry point of the area to be covered Cover all target areas in sequence, and finally return to the task end point P after covering all target areas end , connect the optimal coverage paths of each target area in sequence to obtain the overall optimal coverage path; Step 1.3: Calculate the total length of the path based on the overall optimal coverage path, and use the inverse of the total length of the path as the fitness value.

6. The multi-agent collaborative detection multi-area coverage task planning method based on the improved discrete grey wolf optimization algorithm according to claim 1 is characterized in that: In step 5, the calculation formula of s is: Where iter is the current iteration number, and u is the preset control factor value.

Images