Formation optimization method for multi-UAV ground search mission under implicit communication conditions

Abstract

The present invention discloses a method for optimizing the formation of a multi-UAV ground search mission under implicit communication conditions. The present invention takes the relative position relationship between UAVs as the decision variable, the topological performance index and the mission efficiency index as the objective function, and the size of the target area, the connectivity of the information transmission link, the sensor field of view, etc. as constraints. A formation structure optimization problem model is established, and a multi-objective optimization algorithm is applied to solve it, and finally the desired formation of the UAV formation in the ground target search mission can be obtained. The present invention takes into account the search mission efficiency and topological structure performance at the same time, and is suitable for situations where UAV formations of different sizes perform ground search missions.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-UAV (Unmanned Aerial Vehicle, UAV) formation optimization, and in particular to a multi-UAV ground search mission formation optimization method under implicit communication conditions. Background Art

[0002] With the continuous improvement of UAV technology, UAV has gradually become an important tool for searching and reconnaissance of target areas. Due to the complex and changeable environment and the strong mobility of the target, it is difficult for a single UAV to perform a search mission. The collaborative search mission of multiple UAVs can improve the mission efficiency and achieve complementary advantages. Therefore, multi-UAV collaborative search has also been more widely used. Most literatures study multi-UAV collaborative search under explicit communication conditions, that is, UAVs use communication media and communication equipment to transmit information. However, when UAVs enter a strong electromagnetic interference area, commonly used explicit communication methods such as electromagnetic communication fail. At this time, UAVs can use implicit communication to establish an information transmission link, that is, each UAV detects and transmits information through a limited field of view sensor. At present, there is little research on UAV collaborative search methods under implicit communication conditions.

[0003] To ensure the connectivity of the UAV information transmission link under implicit communication conditions, UAVs can be ordered to search for ground targets in formation. Since implicit communication cannot transmit the structural information of the formation, the formation of the UAV formation needs to be optimized offline, and the target formation information is loaded into the UAV before executing the mission. The optimization of the formation structure is mainly divided into two types: geometric structure optimization and information transmission topological structure optimization. Most literatures study the optimization of the two structures separately, but there is a mapping relationship between the geometric structure and the topological structure, and the changes of the two structures will affect each other. Therefore, it is more practical to combine the two for analysis. Summary of the invention

[0004] In view of this, the present invention provides a method for optimizing the formation of a multi-UAV ground search mission under implicit communication conditions, while ensuring the optimization of the geometric structure of the formation and the optimization of the information transmission topology structure, so as to obtain a better formation.

[0005] The technical solution for implementing the present invention is as follows:

[0006] Step 1: Determine the decision variables of the formation optimization model, wherein the decision variables include the Euclidean distance and the relative sight angle between the UAVs, and the relative sight angle is the angle between the line connecting the two UAVs and the positive direction of the horizontal axis of the geodetic coordinate system.

[0007] Step 2: Determine the constraints of the formation optimization model, which specifically include three constraints:

[0008] (1) Target area constraint: Each UAV is required to perform its mission within the target area.

[0009] (2) Minimum distance constraint between UAVs: To avoid collisions between UAVs, and considering that a close distance between UAVs will have a greater impact on each other’s movement, it is stipulated that the distance between UAVs should not be less than the set minimum distance.

[0010] (3) Connectivity constraint of the information transfer topology graph: A sufficient condition for a UAV cluster to achieve consistency when moving forward is that the cluster’s information transfer topology graph is connected.

[0011] Step 3: Determine the evaluation function of the formation optimization model, which specifically includes two evaluation indicators:

[0012] (1) Indestructibility of information transmission topology structure: The indestructibility of information transmission topology structure means that after some intelligent agents in the cluster are damaged or fail, the remaining intelligent agents can still maintain the connectivity of the information transmission topology graph. The present invention uses natural connectivity to evaluate the indestructibility of information transmission topology structure. Natural connectivity describes the number of closed-loop paths of different step lengths in the network in physics (paths that start from a node and pass through several nodes and still return to the node). The larger the natural connectivity, the more closed-loop paths there are, the more redundant and alternative paths there are between nodes, and the stronger the indestructibility of the topology structure.

[0013] (2) Search mission efficiency: The present invention uses the lateral search coverage width of the UAV formation to evaluate the efficiency of the search mission. The larger the lateral coverage width, the larger the area searched along the specified path, which can improve the area coverage rate.

[0014] Step 4: Integrate the contents of steps 1 to 3 to obtain the formation optimization problem model.

[0015] Step 5, applying a multi-objective optimization algorithm to solve the above problem, and the optimal formation can be obtained. The multi-objective optimization algorithm can adopt NSGA-II-differential evolution joint algorithm, decomposition-based multi-objective optimization algorithm (MOEA / D), multi-objective particle swarm optimization algorithm (MOPSO) and other algorithms.

[0016] Preferably, the present invention combines NSGA-II algorithm and differential evolution algorithm to solve the above problem.

[0017] Preferably, the solution method combining the NSGA-II algorithm and the differential evolution algorithm is as follows:

[0018] S1, first determine the encoding form of the solution, initialize the population P, set the number of solutions in the population, the maximum number of iterations, the maximum distance between nodes, and the minimum distance;

[0019] S2, apply the crossover and mutation operators of the differential evolution algorithm to operate on population P to obtain a new population Q, where the number of solutions in Q is the same as that in P;

[0020] S3, decoding the solutions in population P and population Q to obtain the geometric structure and information transmission topology of the formation; judging whether it is a feasible solution based on the constraint conditions, and if it is a feasible solution, calculating the corresponding evaluation function value; if it is an infeasible solution, calculating its constraint violation amount;

[0021] S4, merge population P and population Q into a new population T, and perform non-dominated sorting on the feasible solutions in population T.

[0022] S5, for feasible solutions with the same non-dominated level, calculate the congestion degree of each solution.

[0023] S6, select a specified number of solutions from population T and use these solutions to update population P; wherein, firstly, the feasible solutions in population T are added to the new population P′ in the order of non-dominated level from small to large, and the feasible solutions with the same non-dominated level are added to population P′ in the order of crowdedness from large to small, and then the infeasible solutions in population T are added to population P′ in the order of constraint violation from small to large; the size of population P′ is equal to that of population P, and the new population P′ is updated to population P;

[0024] Repeat S2 to S6 and stop when the maximum number of iterations is reached to obtain the Parato solution set of the optimization problem;

[0025] S7, in order to ensure the invulnerability of the formation under implicit communication conditions, the solution with the highest invulnerability evaluation function value is selected as the final solution in the Parato solution set.

[0026] Preferably, in S3, the method for acquiring the geometric structure of the formation is as follows:

[0027] Any two drones except the overall pilot drone v i and v j , according to the v obtained from the decoded content i and v j The Euclidean distance and relative sight angle between the pilot drone v1 and the position vector and Then we can get the vector by vector subtraction So we get v i and v j The Euclidean distance between ij and relative sight angle

[0028] Preferably, the connectivity constraint of the drone information transmission topology structure is: the adjacency matrix A of the directed graph D of the drone information transmission topology structure contains only one all-zero column, which corresponds to the overall pilot drone.

[0029] Preferably, the natural connectivity The calculation formula is as follows:

[0030]

[0031] Among them, S is the total number of closed-loop paths in the undirected graph of the drone information transmission topology structure; N is the total number of drones, λ i It is the eigenvalue of the adjacency matrix of the undirected graph G of the drone information transfer topology.

[0032] The optimal efficiency evaluation function f2 of the search mission of the UAV formation is:

[0033]

[0034] Among them, d range is the lateral search coverage width of the UAV formation, N is the total number of drones, and R is the ground search range of the drones.

[0035] Beneficial effects:

[0036] (1) The present invention uses the Euclidean distance and relative line of sight angle between UAVs as decision variables, the limited detection range of drone sensors, the size of the target area, the shortest distance between UAVs (to prevent collisions), the connectivity of the information transmission topology graph, etc. as constraints, and uses the information transmission topology performance index of the formation and the search task efficiency as objective functions to establish a formation optimization model. The model simultaneously considers the geometric structure optimization and information transmission topology optimization of the formation, as well as the search task efficiency and topological structure performance, and is suitable for situations where UAV formations of different sizes perform ground search tasks. The formation optimization model can be solved by using a multi-objective optimization algorithm and is universal.

[0037] (2) The present invention solves the multi-UAV ground search mission formation optimization model based on the NSGA-II-differential evolution joint algorithm. The algorithm is better at solving low-dimensional multi-objective optimization problems and can maintain the diversity of solutions in the population.

[0038] (3) The relative position relationship between each intelligent agent in the formation is solved through vector calculation, which reduces the redundancy of coding and the complexity of calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is the forward detection range of the UAV sensor.

[0040] Figure 2 It is a connected directed tree topology (N=5).

[0041] Figure 3 is the adjacency matrix of a connected directed tree topology (N=5).

[0042] Figure 4 For a directed graph ( Figure 4 (a)) is transformed into an undirected graph ( Figure 4 (b)) in the closed path.

[0043] Figure 5 Schematic diagram of the horizontal coverage width of the search task.

[0044] Figure 6 Flowchart for solving the lateral coverage width for the search task.

[0045] Figure 7 Flowchart of the multi-objective optimization method based on NSGA-II algorithm.

[0046] Figure 8 Schematic diagram of the distribution of evaluation function values ​​of solutions in a certain level set.

[0047] Fig. 9 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0048] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0049] The present invention provides a formation optimization method for a multi-UAV ground search mission under implicit communication conditions. The relative position relationship between UAVs is used as a decision variable, the topological performance index and the mission efficiency index are used as the objective function, and the size of the target area, the connectivity of the information transmission link, the sensor field of view, etc. are used as constraints. A formation structure optimization problem model is established and a multi-objective optimization algorithm is applied to solve it. Finally, the desired formation of the UAV formation in the ground target search mission can be obtained. The flow chart is shown in FIG. Fig. 9 shown.

[0050] Assume that there are N UAVs in the drone cluster and all UAVs are at the same altitude. Therefore, consider the UAV formation in a two-dimensional plane and define the size X max ×Y max (km 2 ) target area. Figure 1The forward detection range of the sensors carried by the UAV is shown. The maximum detection distance is dis, and the forward viewing angle (the angle between the detection range boundary and the UAV direction) is ±ang (ang∈(0,π / 2)). The UAV can only obtain information about other UAVs within its own detection range. The detection range constraint determines that information can only be transmitted one-way between UAVs. Therefore, a directed graph D=(V,E,A) is used to represent the information transmission topology of the UAV cluster, where V={v1,v2,...,v N} is the set of all UAVs, E is the edge set of the directed graph, if v i v can be obtained by detection j information (i, j≤N), then there exists a path from v j Point to v i Directed edge <v j , v i >∈E. A=(a ij ) N×N is the adjacency matrix of the directed graph D. If there are directed edges <v j , v i >∈E, then a ji =1, otherwise a ji =0.

[0051] Problem Modeling:

[0052] Step 1: Identify the decision variables for the optimization problem.

[0053] The decision variables of the formation optimization problem of the present invention include the Euclidean distance between every two nodes and the relative sight angle (the angle between the line connecting the two points and the positive direction of the horizontal axis of the earth coordinate system). i and v j The Euclidean distance between (i≠j) is denoted as l ij , v j Relative to v i The sight angle is denoted as The above decision variables are stored in the matrix L = (l ij ) N×N and Among them, l ii =0,

[0054] Step 2: Determine the constraints of the optimization problem.

[0055] (1) Target area constraints

[0056] Assume the absolute position coordinates of UAV is (X i , Y i ), i = 1, ..., N, each UAV is required to perform tasks in the target area, then

[0057] 0≤Xi ≤X max , 0≤Y i ≤Y max

[0058] (2) Minimum distance constraint between UAVs

[0059] In order to avoid collisions between UAVs, and considering that the movement of each other will be greatly affected when the distance between UAVs is close, it is stipulated that the distance between UAVs should not be less than the set minimum distance d. min ,Right now

[0060] l ij ≥d min , i≠j, i=1,...,N, j=1,...,N

[0061] (3) Connectivity constraints

[0062] The sufficient condition for the UAV cluster to achieve consistency when moving forward is that the topological structure graph of the cluster's information transmission is connected. i You can get node v j The information must meet the following conditions: Let the orientation angle of all UAVs (the angle between the orientation and the positive direction of the horizontal axis of the geodetic coordinate system) be α, then

[0063] l ij ≤dis

[0064]

[0065] According to the above conditions, the adjacency matrix A of the information transmission topology graph (directed graph) can be obtained.

[0066] The topology of the model is a directed tree topology, such as Figure 2 As shown in Figure 1, the arrowed lines represent directed edges in the information transmission topology graph D. In A, only the column corresponding to the general navigator of the cluster is full of 0 columns, indicating that all UAVs except the general navigator can detect the information of UAVs in the cluster. The adjacency matrix (N=5) of the connected directed tree topology is as follows: Figure 3 As shown in the figure, only the third column is all zero, and the corresponding UAV is the overall leader of the cluster. If there are other all-zero columns, it means that the UAV corresponding to the column is out of the cluster and the topology graph is not connected. Let A contain lea all-zero columns, then the connectivity constraint is expressed as:

[0067] lea=1

[0068] That is, there can only be one all-zero column in A, which corresponds to the chief navigator of the UAV cluster, and there can only be one chief navigator in the formation.

[0069] Step 3: Determine the evaluation function of the optimization problem.

[0070] (1) Topological structure indestructibility: Topological structure indestructibility means that after some intelligent agents in the cluster are damaged or fail, the remaining intelligent agents can still maintain the connectivity of the information transmission topology graph.

[0071] The present invention uses natural connectivity to evaluate the invulnerability of topological structures. Natural connectivity describes the number of closed-loop paths of different step lengths in the network in physics, and describes the network's invulnerability from the perspective of redundant alternative paths. A closed-loop path refers to a path that starts from a node and passes through several nodes and still returns to the node. Let n k is the number of closed-loop paths with a path length of k in the network, then the total number of closed-loop paths S in the network is

[0072]

[0073] The longer the closed-loop path is, the greater the communication cost of the node is, and the working performance is relatively poor. Therefore, the introduction of n in the above formula k To prevent the divergence of S, the weights are designed as follows:

[0074]

[0075] Let λ i is the eigenvalue of the adjacency matrix, then

[0076]

[0077]

[0078] When N is large, S is also large, so the above formula is simplified to obtain the natural connectivity The calculation formula is as follows:

[0079]

[0080] The topology considered in this model is a directed tree topology. The number of redundant replacement paths from the root node to the child node can reflect the network's anti-destruction ability. The more redundant paths there are, the more likely the child node can obtain the root node's information through the redundant path when a node between the root node and the child node is destroyed. However, there is no such closed path in the directed tree topology, and natural connectivity cannot be directly applied. The directed tree topology is a weakly connected structure. Ignoring the direction of the edges in graph D, we can obtain an undirected graph G = (V, E', A'). In graph G, multiple paths from the root node to the child node can form a closed path, such as Figure 4 shown.

[0081] The more such closed paths there are, the more redundant replacement paths there are from the root node to the child node, and the better the network's anti-destruction performance. Therefore, according to the eigenvalue of the adjacency matrix A' of the graph G and the natural connectivity calculation formula, the natural connectivity of the graph G can be obtained: Used to characterize the invulnerability of the information transmission topology D. The larger the value of, the stronger the topological structure's invulnerability.

[0082] When graph G is a fully connected graph, it has the maximum natural connectivity The invulnerability evaluation function is obtained through normalization:

[0083]

[0084] (2) Search mission efficiency: The present invention applies the search lateral coverage width d of the UAV formation range The efficiency of the search task is evaluated by . The larger the lateral coverage width, the larger the area searched along the specified path, which can improve the area coverage rate. Assuming that the ground detection area of ​​the UAV is a circular area with a radius of R, the lateral coverage width of the search task is as follows: Figure 5 As shown, the calculation method is as Figure 6 shown.

[0085] Right range Normalization is performed to obtain the search task efficiency evaluation function:

[0086]

[0087] in

[0088]

[0089] When the lateral width of the coverage area is However, due to the forward viewing angle ang<π / 2, the UAV cluster cannot be arranged in a horizontal line.

[0090] Step 4: Integrate the contents of steps 1-3 to obtain the formation optimization problem model.

[0091] maxf(x)=(f1,f2) T

[0092]

[0093] Step 5, problem solving:

[0094] A multi-objective optimization algorithm is applied to solve the above problems.

[0095] First, the following definitions are introduced:

[0096] Parato dominance: For the above objective function value maximization problem, the solution x a dominatex b , if and only if (f1(x a )>f1(x b ))∧(f2(xa )≥f2(x b )) or (f1(x a )≥f1(x b ))∧(f2(x a )>f2(x b )). “∧” represents the AND operation.

[0097] Parato optimal solution: feasible solution x * It is called a Parato optimal solution if and only if there is no solution that can dominate x * The set of all Parato optimal solutions is called the Pareto solution set.

[0098] The process of solving the algorithm is as follows Figure 7 As shown. First, the crossover and mutation operators of the differential evolution algorithm are applied to generate a new population Q based on the original population P, and the two populations are merged to obtain a new population T. Each solution in T has two attributes: non-dominated level (indicating the number of other solutions that can dominate the solution) and crowding (indicating the density of the solution's evaluation function value arrangement). First, the solutions in T are sorted by non-dominated level. This step ensures that the retained solutions are at a better non-dominated level. Then, the solutions with the same non-dominated level are sorted by crowding, and solutions with higher crowding are retained. This step can ensure the diversity of solutions in the population. Finally, some solutions are selected to enter the next generation population based on the solution's non-dominated level, crowding, and population capacity.

[0099] The specific steps include the following:

[0100] Step 5.1: Initialize the population.

[0101] (1) Determine the encoding format of the solution

[0102] The solution code includes the Euclidean distance and relative line of sight angle between UAV No. 1 (the overall pilot drone) and the remaining UAVs, as shown below:

[0103]

[0104] (2) Specify the number of solutions in the population M and the maximum number of iterations G max , the maximum distance dmax and the minimum distance dmin between nodes.

[0105] (3) Initialize the population P

[0106] For the part x in the solution that represents the Euclidean distance between UAV No. 1 and the other UAVs l =(l 12 , ..., l 1N ), the initialization method is as follows, where rand1 is a random number uniformly distributed in [0, 1].

[0107] l 1i =dmin+rand1×(dmax-dmin), i=2,...,N

[0108] For the part of the solution that represents the relative sight angle between UAV No. 1 and the other UAVs, The initialization method is as follows, where rand2 is a random number uniformly distributed in [0, 1).

[0109]

[0110] Step 5.2: For population P, apply the crossover and mutation operators of the differential evolution algorithm to generate a new population Q, where the number of solutions in Q is the same as that in P.

[0111] (1) Variation

[0112] For the solution x in population P i Two parts of (i≤M) and Perform mutation operations respectively and obtain the mutation vector v through mutation i .

[0113]

[0114]

[0115]

[0116] F is the mutation factor, F∈[0,2]. In order to make the algorithm have good global search ability and local search ability, F is decreased as the number of iterations increases. The expression is as follows, where G is the current number of iterations.

[0117]

[0118] (2) Cross

[0119] The solution in the population P is cross-operated with the mutation vector obtained by mutation. The method is as follows:

[0120]

[0121] Where rand is a random number uniformly distributed within [0, 1], x i.j , v i.j x represents the solution i , v i For the j-dimensional data, CR is the crossover factor, CR∈[0.1, 0.9]. The value of CR decreases as the number of iterations increases. The expression is:

[0122]

[0123] (3) Correct the newly obtained solution q

[0124] Since l∈[dmin,dmax], Then make the following corrections to the solution obtained after mutation and crossover:

[0125] q i.j =dmax, if q i.j >dmax.i≤M,1≤j≤N-1

[0126] q i.j =dmin, if q i.j <dmin.i≤M,1≤j≤N-1

[0127]

[0128]

[0129] Where mod(·) is the remainder calculation. The solution in population Q is obtained by correction.

[0130] Step 5.3: Decode the solutions in population P and population Q to determine whether they are feasible solutions. If so, calculate the corresponding evaluation function value.

[0131] (1) Decoding to obtain the geometric and topological structures of the formation

[0132] i) Geometry: Consider any two UAVs except UAV 1 (set as v i and v j ), v can be obtained in the decoded content i and v j The Euclidean distance and relative sight angle between UAV No. 1 are used to obtain the position vector and Vector Subtraction So we get v i and v j The Euclidean distance between ij and relative sight angle The specific calculation method is shown in the following formula:

[0133]

[0134]

[0135]

[0136] Applying the above method, we can obtain the matrix L, φ, which is the Euclidean distance and relative line of sight angle between all UAVs.

[0137] ii) Topology: v i and v j If satisfied:

[0138] l ij ≤dis

[0139]

[0140] Where α is the heading angle (the angle between the AUV heading and the positive horizontal axis of the geodetic coordinate system). Then v j In v i Within the detectable range, there is a path from v j Point to v i The directed edges of the formation are obtained through the above criteria.

[0141] (2) Feasible solution judgment

[0142] Determine whether a solution is a feasible solution based on the constraints in step 2. If the solution satisfies all constraints, it is a feasible solution, otherwise it is an infeasible solution.

[0143] (3) Calculate the evaluation function value

[0144] i) For a feasible solution: Calculate the evaluation function value according to the formula in step 3.

[0145] ii) For an infeasible solution: calculate its constraint violation amount. If the solution violates m constraints, then its constraint violation amount is m.

[0146] Step 5.4: Merge population P and population Q to obtain a new population T, and perform non-dominated sorting on the solutions in population T.

[0147] (1) For each solution in population T, record the number of solutions that can dominate it as n, and the initial value of the non-dominated rank value of each solution is +∞.

[0148] (2) Add the solutions whose level value is +∞ and n is 0 to the set F t (the initial value of t is 1), F t The level of the intermediate solution is t. Let t=t+1.

[0149] (3) Select solutions whose level values ​​are not +∞ from the population T, and perform the following operation on the remaining solutions: n = n-1 for each solution.

[0150] (4) Repeat operations (2) and (3) in step 5.4 until the rank values ​​of all solutions are not +∞, and obtain the non-dominated rank order of all solutions in T.

[0151] Step 5.5: For each set with the same non-dominated level obtained in step 5.4, calculate the congestion degree of all solutions in the level set.

[0152] The arrangement of solutions in a non-dominated level set in the target space is as follows: Figure 8 As shown, the horizontal and vertical axes represent two evaluation functions f1 and f2 respectively. The greater the crowdedness of the solution, the more sparsely populated the solution is in the target space. Retaining such a solution is conducive to improving the diversity of solutions in the population. Conversely, the smaller the crowdedness of the solution, the more densely populated the solution is in the target space. Not retaining such a solution is conducive to improving the speed of the algorithm to obtain the Parato set.

[0153] If the f1 or f2 of a solution is the maximum value of the level set, the crowding degree of the solution is +∞, which means that if the solution in the level set is selected to enter the next generation population, the solution with the maximum value of f1 or f2 can definitely be retained.

[0154] For solutions where the evaluation function value is not an extreme value:

[0155] (1) For a set of feasible solutions: Let a solution be x i , and its corresponding objective function value is and Find the two solutions x before and after each evaluation function direction i+1 and x i-1 , then solve x i The congestion degree is:

[0156]

[0157] in is the extreme value of the objective function.

[0158] (2) For the set of infeasible solutions: the congestion degree is not calculated.

[0159] Step 5.6: Select a specified number of solutions from the population T and update the population P with these solutions.

[0160] According to the non-dominated level value of the solution, it is added to the new population P' in order from small to large. When the addition of a solution in a non-dominated level set makes the number of solutions in P' exceed the limit (i.e. the size of the population P), if the set is a set of feasible solutions, then the solution is added to P' according to the crowdedness of the solution in the set from large to small; if it is a set of infeasible solutions, then the solution is added to P' according to the constraint violation amount from small to large. Finally, let P = P'. When adding solutions to P', add feasible solutions first, and then add infeasible solutions.

[0161] Step 5.7: Repeat steps 5.2 to 5.6, and stop when the number of iterations equals the maximum number of iterations, and finally obtain the Parato solution set.

[0162] Step 5.8: To ensure the stability of the formation, select the solution with the highest value of the invulnerability evaluation function in the Parato solution set as the final solution.

[0163] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A formation optimization method for multi-UAV ground search mission formation under implicit communication conditions, characterized in that: The Euclidean distance and relative line of sight angle between UAVs are used as decision variables of the formation optimization model. The target area constraint, the shortest distance constraint between UAVs, and the connectivity constraint of the UAV information transmission topology structure are used as constraint conditions of the formation optimization model. The indestructibility of the UAV information transmission topology structure and the efficiency of the search task of the UAV formation are used as evaluation functions of the formation optimization model. A formation optimization model for multi-UAV ground search tasks is established. First, the UAV information transmission topology structure is converted into an undirected graph, and then the natural connectivity is used to evaluate the indestructibility of the UAV information transmission topology structure. The lateral search coverage width of the UAV formation is used to evaluate the efficiency of the search task. The connectivity constraint of the drone information transmission topology structure is: the adjacency matrix A of the directed graph D of the drone information transmission topology structure contains only one all-zero column, and the all-zero column corresponds to the total pilot drone; The natural connectivity The calculation formula is as follows: Among them, S is the total number of closed-loop paths in the undirected graph of the drone information transmission topology structure; N is the total number of drones, λ i is the eigenvalue of the adjacency matrix of the undirected graph G of the drone information transfer topology; The invulnerability evaluation function f1 of the UAV information transmission topology is: in, The maximum natural connectivity of the undirected graph G of the human-computer information transmission topology; The efficiency evaluation function f2 of the search mission of the UAV formation is: Among them, d rande is the lateral search coverage width of the UAV formation, N is the total number of drones, and R is the ground search range of the drones; The multi-UAV ground search mission formation optimization model is solved to obtain the optimal formation.

2. The formation optimization method for multi-UAV ground search mission formation under implicit communication conditions as claimed in claim 1, characterized in that: The multi-UAV ground search mission formation optimization model is solved based on the NSGA-II-differential evolution joint algorithm, the decomposition-based multi-objective optimization algorithm or the multi-objective particle swarm optimization algorithm.

3. The formation optimization method for multi-UAV ground search mission formation under implicit communication conditions as claimed in claim 2, characterized in that: The specific steps of solving the problem based on the NSGA-II-differential evolution joint algorithm are as follows: S1, determine the encoding form of the solution including the Euclidean distance and relative line of sight angle between the overall pilot UAV and the remaining UAVs; determine the number of solutions in the population, the maximum number of iterations, the maximum distance and the minimum distance between UAVs; initialize the population P; S2, use the crossover and mutation operators of the differential evolution algorithm to operate on the population P to obtain a new population Q, in which the number of solutions is the same as that of P; S3, decoding the solutions in population P and population Q to obtain the geometric structure and information transmission topology of the formation; judging whether the solution is a feasible solution based on the constraint conditions, and if it is a feasible solution, calculating the corresponding evaluation function value; if it is an infeasible solution, calculating the constraint violation amount; S4, merge population P and population Q into a new population T, and perform non-dominated ranking on all feasible solutions in population T; S5, for feasible solutions with the same non-dominated level, calculate their congestion degree; S6, add the feasible solutions in population T to the new population P' in order of non-dominated level from small to large, and add the feasible solutions with the same non-dominated level to the population P' in order of crowdedness from large to small; if the scale of feasible solutions in population T is smaller than the scale of population P, continue to add the infeasible solutions in population T to the population P' in order of constraint violation from small to large until the scale of population P' is equal to that of population P; update the new population P' to population P, return to S2, until the number of iterations is equal to the maximum number of iterations, and obtain the Parato solution set of the optimization problem; S7, select the solution with the highest value of the invulnerability evaluation function in the Parato solution set as the final solution.

4. The formation optimization method for multi-UAV ground search mission formation under implicit communication conditions as claimed in claim 3, characterized in that: In S3, the method for obtaining the geometric structure of the formation is as follows: Any two drones except the overall pilot drone v i and v j , according to the v obtained from the decoded content i and v j The Euclidean distance and relative sight angle between the pilot drone v1 and the position vector and Then we can get the vector by vector subtraction So we get v i and v j The Euclidean distance between ij and relative sight angle

Images