Connection decision method for unmanned cluster generation considering resource replenishment

Abstract

This invention discloses a connection decision-making method for unmanned swarm generation considering resource replenishment, comprising the following steps: obtaining the swarm structure entropy under the current swarm size and the swarm size after introducing the unmanned system, and constructing a value function; based on the value function, allocating the task types of the introduced unmanned system to obtain the number of decision-making nodes, perception nodes, and execution nodes; constructing a preference connection model, and sequentially introducing decision-making nodes, perception nodes, and execution nodes based on the preference connection model to complete the resource replenishment of the swarm; and optimizing the connection state of the swarm under resource replenishment using a particle swarm optimization algorithm. The connection decision-making method for unmanned swarm generation considering resource replenishment proposed in this invention can ensure the reliable operation of the unmanned swarm.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) swarm control technology, and particularly relates to a UAV swarm generation and connection decision method that takes into account resource replenishment. Background Technology

[0002] An unmanned swarm system refers to a system in which several unmanned systems collaborate to complete complex tasks within a certain time and space, based on task division. With the increasing intelligence and complexity of task scenarios in unmanned swarm systems, the system typically dynamically supplements or replaces unmanned systems when tasks are lost, continuously withdrawing and replenishing resources according to task requirements. In other words, the system structure is open, and its composition can be adjusted as needed, eliminating some old systems and absorbing new ones. During this process, task reliability exhibits a high degree of uncertainty and instability. Therefore, there is an urgent need to propose an unmanned swarm generation and connection decision-making method that considers resource replenishment to efficiently and rationally utilize resources, thereby ensuring the high reliability of the unmanned swarm operation. Summary of the Invention

[0003] This invention proposes a connection decision method for unmanned cluster generation that considers resource replenishment, in order to solve the technical problems existing in the prior art.

[0004] To achieve the above objectives, this invention provides a method for unmanned cluster generation connection decision considering resource replenishment, comprising the following steps:

[0005] Obtain the cluster structure entropy under the current cluster size and the cluster size after introducing the unmanned system, and construct the value function.

[0006] Based on the value function, the task types of the introduced unmanned system are allocated to obtain the number of decision-making nodes, the number of perception nodes, and the number of execution nodes.

[0007] A preference connection model is constructed, and decision-making nodes, perception nodes, and execution nodes are introduced in sequence based on the preference connection model to complete the resource supplementation of the cluster;

[0008] The connection state of the cluster is optimized based on the particle swarm optimization algorithm under resource replenishment.

[0009] Optionally, the value function J(m) is as follows:

[0010] J(m)=aQ(n+m)-b(S(n+m)-S(n))

[0011] Where Q(n+m) is the performance of a cluster of size n+m, S(n) is the structural entropy of a cluster of size n, S(n+m) is the structural entropy of a cluster of size n+m, n is the current cluster size, n+m is the cluster size after the introduction of the unmanned system, a is the weight parameter for performance, and b is the weight parameter for cost.

[0012] Alternatively, the preference connection model is as follows:

[0013]

[0014] Among them, Π(k) i Let k be the probability that a new node is connected to node i. i Let η be the degree of node i. i Let i be the fitness degree of node i.

[0015] Optionally, the process of introducing decision-making nodes includes: establishing full connections between the introduced decision-making nodes and the remaining decision-making nodes in the cluster.

[0016] Optionally, the process of introducing a perception node includes: finding the node with the smallest degree among decision-making nodes in the cluster and making a full connection with the introduced perception node; if the degree of decision-making nodes in the cluster is the same, then selecting the nearest decision-making node and making a full connection with the introduced perception node; if the connections of all decision-making nodes have reached the communication limit, then selecting the node with the smallest degree among the remaining perception nodes in the cluster and making a full connection with the introduced perception node.

[0017] Optionally, the process of introducing execution-type nodes includes: finding the node with the smallest degree among decision-type nodes in the cluster and making a full connection with the introduced execution-type node; if the degree of decision-type nodes in the cluster is the same, then selecting the nearest decision-type node and making a full connection with the introduced execution-type node; if the connections of all decision-type nodes have reached the communication limit, then selecting the node with the smallest degree among the remaining execution-type nodes in the cluster and making a full connection with the introduced execution-type node.

[0018] Optionally, the process of replenishing cluster resources also includes: continuously introducing decision-making nodes, perception nodes, and execution nodes in the order of first looping nodes and then looping edges, until all nodes in the cluster reach the communication capacity limit, terminating the loop and completing the replenishment of cluster resources.

[0019] Optionally, the process of optimizing the connection state of the cluster under resource supplementation based on the particle swarm optimization algorithm includes: initializing the scale of the particle swarm based on the scale of the introduced unmanned system, and randomly assigning the position and velocity of each particle; then introducing the first node, calculating the fitness of each particle, updating the individual's best historical position and the global best historical position based on the fitness, and updating the velocity and position of each particle; repeating the update steps until the maximum number of iterations is reached or the optimal position termination condition is reached, then obtaining the optimal state of the first node, updating the entire network, and finally repeating the optimization process to find all remaining nodes.

[0020] Compared with the prior art, the present invention has the following advantages and technical effects:

[0021] The unmanned cluster generation connection decision method proposed in this invention takes into account resource replenishment. By measuring the recovery cost of the cluster and making a decision based on the comprehensive performance improvement and cost, it can ensure the reliable operation of the unmanned cluster. Attached Figure Description

[0022] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0023] Figure 1 This is a flowchart illustrating the resource replenishment process under the cluster preference connection rules in this embodiment of the invention.

[0024] Figure 2 This is a flowchart illustrating the cluster resource supplementation process based on the particle swarm optimization algorithm according to an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of the curve of J(m) versus m when the scale is 98 in this embodiment of the invention;

[0026] Figure 4 This is a schematic diagram of the curve of J(m) as a function of m when the scale is 101 in this embodiment of the invention;

[0027] Figure 5 This is a schematic diagram of the curve of J(m) changing with m when the scale is 104 in an embodiment of the present invention. Detailed Implementation

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0030] Example 1

[0031] This embodiment provides a connection decision method for unmanned swarm generation that considers resource replenishment. In order to understand the multi-layered coupled unmanned swarm system applied in complex scenarios, the system is first modeled based on its multi-dimensional characteristics. Using complex network modeling methods, the unmanned systems in the system are treated as nodes, and the interactions between them are considered as edges, forming a complex network. Furthermore, based on the composition of the unmanned system components, the multi-layered complex network is divided into a communication layer, a structural layer, and a task layer, corresponding to the communication data link, unmanned platform, and task payload of the unmanned swarm system, respectively, thereby decoupling the complex relationships between the components in the swarm.

[0032] Based on the multidimensional coupling performance of clusters and considering their specific applications in various scenarios, a performance index system for unmanned clusters under resource supplementation was established. This includes structural topology performance: measuring the cluster's formation topology in space; traffic connectivity performance: measuring the flow of cluster communication connections and task allocation; and task effectiveness: measuring the cluster's performance in executing tasks.

[0033] Based on this, and according to the relationship between the potential and performance of unmanned swarms, a potential health index is established using the concept of reliability entropy to measure the quality of performance in relation to the above three performance aspects. The swarm's potential is measured by the number of potential states that can be formed under the current state, thereby understanding the utilization efficiency of the swarm system.

[0034] A method for replenishing unmanned cluster resources based on preference-based connections:

[0035] Preferred connections are an important concept in complex network theory. By formulating preferred connection strategies, new nodes can connect to existing nodes with a certain probability. Applied to cluster networks, this can enable network reconstruction decisions under resource replenishment. Therefore, the problem of elastic resource replenishment can be broken down into two parts: first, determining the scale of resource introduction, and then using preferred connections to make connection decisions.

[0036] Scale decision based on the balance between resource replenishment performance and recovery costs:

[0037] In resource replenishment mode, cluster recovery is available for every state. To find the optimal state under constraints, the performance improvement and required resources for different scales are first comprehensively weighed. The performance gap between the performance curves of each state is different, and the resource cost required to transition between states also varies. Using the concept of entropy, the state curves can be viewed as equipotential surfaces of electric charge, contour lines of mountains, and state and energy transfer processes such as atomic energy level transitions.

[0038] Entropy represents the directionality of energy transfer; a low-entropy system automatically transforms into a high-entropy system, while the transformation from a high-entropy system to a low-entropy system requires external work. For example, the second law of thermodynamics describes the directionality of entropy in a thermodynamic system. In reliability studies, entropy represents the availability of a system. As a system progresses from its initial state to a degraded state, its entropy continuously increases, its energy availability continuously decreases, and it eventually fails.

[0039] On the other hand, the entropy reduction process requires energy consumption. In a thermodynamic system, the transfer of energy from a low-temperature system to a high-temperature system requires external work because it is an entropy reduction process for a heat transfer system. In a mountain contour map, the higher the altitude, the more difficult it is to rise to the same height. The value of the entropy change represents the cost required for the state change.

[0040] Similarly, in cluster systems, clusters of different sizes have different entropy values. Let the cluster size be n frames. Based on Boltzmann entropy, we can obtain:

[0041]

[0042] in, Let A be the number of states that the cluster can form in state A. n For a cluster of size n frames, This is the formula for calculating permutations and combinations in probability theory, where k is the weighting coefficient, and the difference in entropy between clusters of size n and size m is:

[0043]

[0044] For clusters of adjacent sizes, i.e., let m = n-1, then we have:

[0045]

[0046] In other words, the change in entropy of cluster health status varies with different cluster sizes. The change in entropy is directly proportional to the cluster size; that is, the larger the cluster, the greater the increase / decrease in system availability when the cluster size increases / decreases, i.e., the more available states are provided. Therefore, when supplementing resources, the larger the cluster size, the higher the cost of introducing the required resources.

[0047] When the clusters are of the same size, the entropy of the health state is the same, and the change in the number of potential states caused by different state changes is also the same. For example, the entropy change when state 3 is transferred to state 1 is the same as the entropy change when state 5 is transferred to state 3. In the formula, Ω refers to the number of states that can be formed under the structure of that state.

[0048] ΔS 31 =klnΩ3-klnΩ1=kln2

[0049] ΔS 31 =klnΩ5-klnΩ3=kln2

[0050] Therefore, decisions on replenishing cluster resources can be made by combining changes in health state entropy caused by state changes with performance changes.

[0051] For the introduction of clusters of different sizes, assuming the current size is n and m unmanned systems are introduced, the value function is:

[0052] J(m)=aQ(n+m)-b(S(n+m)-S(n)) (4)

[0053] Where Q(n+m) is the performance of a cluster of size n+m, S(n) is the structural entropy of a cluster of size n, S(n+m) is the structural entropy of a cluster of size n+m, n is the current cluster size, n+m is the cluster size after the introduction of the unmanned system, a is the weight parameter for performance, and b is the weight parameter for cost.

[0054] By optimizing the value function, the scale of introduction can be determined by comprehensively considering factors such as cluster performance improvement and recovery cost constraints. Secondly, the task types of m unmanned systems are allocated, assuming the number of perception, decision-making, and execution nodes introduced are m1, m2, and m3 respectively, with an initial task node ratio of N1, N2, and N3. Then, let m satisfy:

[0055] n1+m1:n2+m2:n3+m3=N1:N2: (5)

[0056] m1 + m2 + m3 = m (6)

[0057] The number of task nodes for each type can be calculated.

[0058] Complex network preference connection mechanism:

[0059] Preferential attachment in complex networks is a probabilistic mechanism that defines the probability of a newly generated node connecting to each existing node. This probability is often related to node attributes. A common type is degree-based preference attachment, where the probability of connecting to node i can be set as follows:

[0060]

[0061] Among them, Π(k) i Let k be the probability that a new node connects to node i. i Let be the degree of node i, indicating the probability of connecting to nodes with higher degree values. Bianconi–Barabási proposed a scale-free model, also known as an adaptive model, where the probability of an edge from a new node connecting to node i is:

[0062]

[0063] Where, η i The fitness of node i is determined by the connection of randomly generated networks according to this preference, which is a scale-free network (adaptive model).

[0064] The mechanism of favored connections can explain why some nodes are more important than others in a network. These nodes, due to their higher connectivity, play a pivotal role in the network, thus increasing their importance. In this model, fitness determines the growth rate of nodes and the growth of the network. When all fitness values ​​are equal, the model simplifies to the Barabási-Albert model, where the degree distribution of a scale-free BA network satisfies the probability p of a node having degree k. k ∝k -3 .

[0065] Connection decision-making method based on preference-based connection rules:

[0066] In the initial state of the cluster, the nodes are connected to form a regular network. During resource replenishment, a degree-based power-law distribution is undesirable, as it can lead to system fragility. Furthermore, when the communication load of an unmanned system is too high, the performance requirements for its equipment are high, making it prone to failure, and network connectivity performance deteriorates. Therefore, during resource replenishment, this embodiment sets a preference connection model opposite to degree-based one, thereby ensuring a more uniform cluster. This is illustrated in the following equation:

[0067]

[0068] That is, when a new node exists, there is a preference to connect to nodes with lower degrees in the current network. The connection rules must satisfy:

[0069] There are no nodes with a degree of 0. Based on the principle of uniformity, each edge is connected in the order of first looping through the nodes, then looping through the edges. That is, the first edge of each node is introduced in turn, then the second edge of each node is introduced, until the end of the loop is reached. The resource replenishment steps can be set as follows:

[0070] (1) Based on the value function J(m), the optimal number of unmanned systems to be introduced is m. Each node is introduced sequentially according to the preference connection rule.

[0071] (2) Introduce decision-making nodes and make full connections between decision-making nodes and other decision-making nodes.

[0072] (3) Introduce perception nodes and determine the node with the smallest decision node degree in the cluster network. If the decision node degrees are the same, select the nearest decision node. If all decision nodes have been connected and the communication limit has been reached, connect the node with the smallest degree among the remaining perception nodes in the formation.

[0073] (4) When introducing execution-type nodes, determine the node with the smallest decision-making node degree in the cluster network. If the decision-making nodes are the same, select the nearest decision-making node. If all decision-making nodes have been connected and the communication limit has been reached, connect the node with the smallest degree among the remaining execution-type nodes in the formation.

[0074] (5) Repeat steps (3) and (4) to connect edges, following the order of first looping through nodes and then looping through edges, until all cluster nodes reach their communication capacity limit, at which point the loop terminates, yielding the network configuration after resource replenishment. The resource replenishment flowchart under the cluster preference connection rule is shown below. Figure 1 As shown.

[0075] Resource replenishment optimization method based on the health status of unmanned clusters:

[0076] The preference-based unmanned cluster resource replenishment method uses the cluster's node degree as the basis for reconstructing connections, but it does not consider the cluster's potential health status during the connection process. To comprehensively explore all cluster states during reconstruction and delve deeper into the state of cluster resource replenishment, this embodiment uses the optimal performance state to make cluster resource replenishment decisions. Using performance as the objective function, the optimal recovery configuration of the cluster is solved to guide the cluster's connection decisions under resource replenishment.

[0077] Analysis of the Particle Swarm Optimization Algorithm:

[0078] Objective optimization methods obtain results by defining a suitable objective function and searching for the optimal solution using a certain search method. Simple optimization problems are often transformed into mathematical programming problems to find exact solutions. However, in real-world scenarios, finding exact solutions usually involves extremely high computational complexity. Therefore, heuristic algorithms have emerged that approximate the optimal solution by searching for the optimum. Based on heuristic algorithms and combining the intelligence of biological individuals and groups, various metaheuristic algorithms have become the mainstream methods for solving optimization problems. Metaheuristic algorithms are often used to solve NP-hard problems. Different algorithms have different application scenarios. This embodiment uses the particle swarm optimization (PSO) algorithm to study state optimization of a cluster under resource replenishment.

[0079] Particle Swarm Optimization (PSO) is an optimization algorithm whose basic idea comes from the foraging behavior of birds. Based on the concept of swarm intelligence, it sets up multiple particles, each with initial parameters such as position and velocity. Each particle searches for its own optimal value, while the swarm provides feedback on the global optimal value. Each particle adjusts its position and velocity based on its individual optimal value and the global optimal value until the iteration ends. The overall process of PSO is as follows:

[0080] (1) Initialize the particle swarm, i.e. set the number of particles, velocity and position, maximum velocity, maximum position, number of iterations and hyperparameters.

[0081] (2) Evaluate the fitness of the particles, that is, calculate the fitness value of each particle according to the objective function.

[0082] (3) Update the individual extreme values ​​and global optimal solutions of particles based on the optimal value of fitness.

[0083] (4) Adjust the velocity and position of the particles based on the velocity update and position update formulas.

[0084]

[0085]

[0086] in, pbest represents the d-dimensional position and velocity vectors of the i-th particle in the k-th iteration. id For the current optimal position of particle i, gbest id Let be the global optimal position of particle i. c1 and c2 are the individual learning factor and the social learning factor, respectively. w, r1, and r2 are the weight coefficients. Repeat steps (2)-(4) until the stopping condition is met, such as reaching the preset maximum number of iterations or finding the optimal solution that meets the accuracy requirements.

[0087] Compared to other metaheuristic algorithms, particle swarm optimization (PSO) is simple to implement and can achieve good results when the problem's conceptual factors are relatively well-defined. Therefore, the decision-making process for reconstructing the cluster's state can be solved by setting a reasonable D-dimensional space and fitness function.

[0088] State optimization based on particle swarm optimization:

[0089] The state optimization problem under resource replenishment in a swarm network is an NP-hard problem. We can define the particle's D-dimensional space as the number of edges that a node can connect to. The position and velocity in this space are discretized, with each discrete point representing a configuration (state) formed by that node. Its velocity represents the orientation of the next position. With the original swarm configuration unchanged, the complexity of this problem is proportional to the exponential power of the number of possible new edges. Considering the excessively high time complexity when the swarm size is increased to m, the problem is addressed step-by-step by adding each new node sequentially. Furthermore, the particle swarm optimization algorithm suffers from getting trapped in local optima; therefore, a perturbation is implemented for the swarm state optimization problem. The improved particle swarm optimization algorithm is as follows:

[0090] (1) The resource replenishment scale is generated through scale decision, i.e., the scale of the introduced unmanned system m. The initial particle swarm size is num. The position and velocity of each particle are randomized. Since the performance of the cluster edges are independent, the distribution should be as uniform as possible within the number of nodes. The position of the i-th particle is x. i =(x i1 ,x i2 ,...x iD ), satisfying 0≤x ij <num(j=1,2,...,D). Velocity is v i =(v i1 ,v i2 ,...,v iD Due to the communication limit, the number of times each location coordinate appears is constrained. Initial hyperparameters w, c1, and c2 are set, and the number of iterations n is set.

[0091] (2) Introduce the first node, calculate the fitness of each particle, update the individual best historical position pbest and the global best historical position gbest according to the fitness, and update the velocity and position of each particle according to the velocity and position update formula. Since the velocity orientation only has three components: inertia, individual best and global best, in order to avoid the algorithm getting trapped in local optima, the particle swarm algorithm is improved by adding a reverse perturbation ε.

[0092]

[0093] Where ε represents a positive perturbation.

[0094] (3) Repeat the above steps until the maximum number of iterations is reached or the optimal position is reached to terminate the process. This indicates the optimal state for the current scale has been found, the optimization of the first point is complete, the entire network is updated, and the above steps are repeated sequentially to find all remaining nodes. The flowchart of the entire process is as follows: Figure 2 .

[0095] Case analysis verification:

[0096] To illustrate the concept of UAV swarm utilization based on state in this embodiment, and to verify the scale decision-making method based on health state, the UAV swarm resource replenishment method based on preference connection, and the optimization method based on swarm health state, the performance recovery of preference connection and proximity connection, and the performance improvement of particle swarm algorithm and traversal algorithm are compared.

[0097] Validation of a scale-based decision-making method that considers both resource replenishment performance and recovery costs:

[0098] The health-state-based scale decision-making method determines the resource replenishment scale by solving the value function J(m), and obtains the number of sensing, decision-making, and execution nodes m1, m2, and m3 through proportional control. This embodiment analyzes and verifies the method through a cluster resource replenishment case.

[0099] The initial cluster size is set to 120, the ratio of perception, decision-making, and execution is 3:1:2, and the cluster response mode is resource replenishment. After a series of tasks, the cluster size decreases. The cluster behavior is simulated using Netlogo software.

[0100] Solve for the changes in the value function J(m) with the resource replenishment number m when the scale is reduced to 98 (perception 54, decision 20, execution 24), 101 (perception 56, decision 20, execution 25), and 104 (perception 59, decision 20, execution 25):

[0101] like Figures 3-5 As shown, J(m) exhibits a trend of first increasing and then decreasing with respect to m, reaching its maximum value at a certain point. When the scale is 98, 101, and 104, the optimal scale and corresponding task type ratios are 12 (perception 0, decision 0, execution 12), 9 (perception 0, decision 0, execution 9), and 6 (perception 0, decision 0, execution 6), respectively, all close to the scale of resource replenishment up to 110. In summary, under different initial scales, the curve of J(m) increasing with the number of nodes varies, but the optimal scale of cluster recovery is less affected by changes in the initial scale. Therefore, as cluster performance increases with resource replenishment, the health state entropy decreases simultaneously, leading to an increase in the cost of cluster resource replenishment. By setting a value function, the scale decision of cluster replenishment can be made by comprehensively considering the performance and health recovery costs, providing a quantitative basis for cluster resource replenishment quantity schemes.

[0102] Validation of the preference-based connection decision method:

[0103] Based on the scale decision results, a case study is conducted to validate the resource replenishment decision when the scale is 98 and the replenishment scale is 12. Both proximity-based and preference-based connection methods are used for reconstruction decisions. The cluster performance curves of the proximity-based and preference-based connection methods are compared to analyze the effectiveness of the preference-based connection state decision method.

[0104] It can be seen that the reconstruction decision mode using preference connection can achieve better performance recovery and also has certain advantages in recovery speed. This indicates that during the state transition process, compared with the nearest connection method, the cluster's reconstruction connection mode can find the best state more quickly and accurately.

[0105] Feasibility analysis of optimization methods based on health status:

[0106] As can be seen from the aforementioned case analysis, resource supplementation connections based on preference connections can effectively improve cluster performance. Therefore, the optimal recovery method for the best state is also feasible. In order to propose a particle swarm optimization algorithm suitable for edge recovery of unmanned clusters, this embodiment conducts a feasibility analysis on the implementation and parameter adjustment of the particle swarm algorithm.

[0107] For the 104-node cluster, resources were supplemented by adding 6 nodes based on the results of the scale decision method. The state after adding 6 nodes was then optimized.

[0108] The initial parameters of the algorithm are set as follows: dimension D = 2, number of particles = 10, number of iterations = 800, maximum position = 104, maximum speed = 40, C1 = 5, C2 = 5, W = 1. The convergence of performance q is as follows: the global optimum value obtained by the traversal algorithm is 4.3844, and the optimal positions are [84, 64], [104, 46], [101, 6], [97, 40], [0, 98], [3, 36], which are the same as the particle swarm optimization algorithm. The convergence speed of the particle swarm optimization algorithm is improved by (1-6808 / 338) = 19 times compared with the traversal algorithm. The original particle swarm optimization algorithm will have local optima, while the improved particle swarm optimization algorithm with perturbation can converge to the optimal value. Therefore, the method of adding perturbation to the cluster can find the global optimum solution, and it is feasible to optimize the cluster based on this method.

[0109] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for unmanned cluster generation connection decision considering resource replenishment, characterized in that, Includes the following steps: Obtain the cluster structure entropy under the current cluster size and the cluster size after introducing the unmanned system, and construct the value function. The cluster structure entropy is: ； in, for The number of states that a cluster can form under a given state. For a cluster of size n frames, This is the formula for calculating permutations and combinations in probability theory. These are the weighting coefficients; Based on the value function, the task types of the introduced unmanned system are allocated to obtain the number of decision-making nodes, the number of perception nodes, and the number of execution nodes. A preference connection model is constructed, and decision-making nodes, perception nodes, and execution nodes are introduced in sequence based on the preference connection model to complete the resource supplementation of the cluster; The connection state of the cluster is optimized based on the particle swarm optimization algorithm under resource replenishment. The value function As shown below: ； in, For scale Cluster performance, Let n be the entropy of the cluster structure. For scale The cluster structure entropy, where n is the current cluster size. The cluster size is determined by the number of unmanned systems introduced, where 'a' is the weight parameter for performance and 'b' is the weight parameter for cost. The preference connectivity model is shown below: ； in, Let be the probability that the new node is connected to node i. Let i be the degree of node i. Let i be the fitness degree of node i; The process of introducing decision-making nodes includes: establishing full connections between the introduced decision-making nodes and the remaining decision-making nodes in the cluster; The process of introducing perception nodes includes: finding the node with the smallest degree among decision-making nodes in the cluster and making a full connection with the introduced perception node; if the degree of decision-making nodes in the cluster is the same, then select the nearest decision-making node and make a full connection with the introduced perception node; if the connections of all decision-making nodes have reached the communication limit, then select the node with the smallest degree among the remaining perception nodes in the cluster and make a full connection with the introduced perception node. The process of introducing execution-type nodes includes: finding the node with the smallest degree among decision-type nodes in the cluster and making a full connection with the introduced execution-type node; if the degree of decision-type nodes in the cluster is the same, then select the nearest decision-type node and make a full connection with the introduced execution-type node; if the connections of all decision-type nodes have reached the communication limit, then select the node with the smallest degree among the remaining execution-type nodes in the cluster and make a full connection with the introduced execution-type node.

2. The unmanned cluster generation connection decision method considering resource replenishment according to claim 1, characterized in that, The process of replenishing cluster resources also includes: continuously introducing decision-making nodes, perception nodes, and execution nodes in the order of first looping nodes and then looping edges, until all nodes in the cluster reach their communication capacity limit, terminating the loop, and completing the replenishment of cluster resources.

3. The unmanned cluster generation connection decision method considering resource replenishment according to claim 1, characterized in that, The process of optimizing the connection state of the cluster under resource replenishment based on the particle swarm optimization algorithm includes: initializing the particle swarm size based on the scale of the introduced unmanned system, and randomly assigning the position and velocity of each particle; then introducing the first node, calculating the fitness of each particle, updating the individual's best historical position and the global best historical position based on the fitness, and updating the velocity and position of each particle; repeating the update steps until the maximum number of iterations is reached or the optimal position termination condition is reached, then obtaining the optimal state of the first node, updating the entire network, and finally repeating the optimization process to find all remaining nodes.

Images