Unmanned aerial vehicle cluster system security assessment method based on multilayer complex network theory

Abstract

The invention provides an unmanned aerial vehicle cluster system security assessment method based on a multi-layer complex network theory, and the method comprises the steps: constructing a multi-layer complex network model, abstracting a task layer and a structure layer of an unmanned aerial vehicle cluster, and considering the dynamic characteristics, various fault forms and task dependence of the unmanned aerial vehicle cluster. Therefore, the complex structure and the interaction relation of the unmanned aerial vehicle cluster system can be described more accurately. Through an improved dynamic propagation model, the propagation process and the recovery process of a fault in the system are simulated, the safety performance of the system is quantitatively evaluated, and the quantitative relation between the system safety and propagation parameters is explored. According to the method, the safety of the unmanned aerial vehicle cluster system can be accurately evaluated, a scientific basis can be provided for optimization management, fault recovery and scientific decision-making of the system, the operation efficiency and reliability of the unmanned aerial vehicle cluster in a complex low-altitude scene are improved, powerful support is provided for wide application of the unmanned aerial vehicle technology, and the method is suitable for popularization and application. And service is provided for the society in a reliable mode.

Description

Technical Field

[0001] This invention provides a method for security assessment of unmanned aerial vehicle (UAV) swarm systems based on multi-layer complex network theory. It relates to a technical implementation of assessing the security of UAV swarm systems based on multi-layer complex network theory and belongs to the field of system reliability. Background Technology

[0002] Currently, drone technology is continuously advancing, with significantly improved autonomy and its application areas gradually expanding from the military to the civilian sector. Drones are increasingly widely used in civilian applications, including but not limited to agricultural monitoring and plant protection, logistics and distribution, environmental monitoring, emergency rescue, urban surveying and planning, and power line inspection. The extensive use of drone swarms has significantly increased the complexity of low-altitude airspace, making the interactions between drones and their interaction with the environment more complex and dynamic. This complexity is reflected not only in the number and types of drones but also in the diversity and dynamism of their missions.

[0003] With the widespread application of drones in the low-altitude airspace, the complexity of low-altitude airspace has increased significantly, mainly in the following three aspects. First, the highly dynamic evolutionary nature: During mission execution, the relative positions and mission states of drone swarms constantly change, making the low-altitude airspace environment more complex. For example, drones may be affected by sudden weather changes (such as thunderstorms or typhoons) or terrain obstacles (such as tall buildings or mountains), leading to mission interruption or failure. Second, the cross-domain propagation of risks: The operation of drone swarms involves multiple domains (such as communication, navigation, and mission execution), and risks can propagate across these domains. For example, a communication link failure may cause drones to lose control, thus affecting their mission execution and flight safety. Third, the concealment and complexity of risks: The propagation of risks is often covert, making it difficult to detect in a timely manner through traditional monitoring methods. For example, implicit interactions between drones (such as electromagnetic interference) may lead to mission failure or flight accidents. At the same time, the propagation path and scope of impact of risks are difficult to predict, increasing the difficulty of risk identification and assessment.

[0004] Current low-altitude models are mainly divided into two categories: data-driven and model-driven. Data-driven modeling suffers from problems such as data dependence and poor adaptability. Current model-driven modeling is either static, failing to consider the dynamic characteristics of UAVs, or abstracts UAV swarms as a single-layer network, making it difficult to accurately describe the multi-layered complex structure of UAV swarm systems and comprehensively reflect the various interactions within the system. Simultaneously, UAV swarm systems involve multi-dimensional risk factors, including the physical characteristics of UAVs, mission characteristics, and environmental factors. Risks in UAV swarm systems are characterized by cross-domain nature, hidden connections, and implicit interactions, making risk identification and assessment more difficult. Traditional risk assessment methods mainly focus on single-dimensional risk factors, making it difficult to identify and assess cross-domain propagation risks and effectively handle the complexity brought about by implicit interactions. Furthermore, the operation of UAV swarms requires consideration of mission dependencies and inter-UAV coordination, which places high demands on real-time decision-making. Existing decision-making methods are often based on static models, making it difficult to adapt to the dynamic changes of UAV swarm systems and unable to make scientifically sound decisions in complex low-altitude scenarios.

[0005] This invention, based on multi-layer complex network theory, abstracts UAV swarms into a two-layer network of structure and task layers. It considers various failure modes of UAV swarms and the dynamic characteristics of UAV structures, while distinguishing UAVs at different task levels. From a systems perspective, it constructs a universal security assessment framework for multi-level UAV swarm systems. This invention can accurately and quantitatively assess the security of UAV swarm systems, thus providing a scientific basis for system optimization management, fault management, and scientific decision-making. It helps to build a universal security assessment framework, supports real-time decision-making, optimizes task allocation and fault recovery strategies, improves the operational reliability of UAV swarms in complex low-altitude scenarios, expands their application in more fields, and simultaneously ensures public safety. Summary of the Invention

[0006] (I) Purpose of the Invention

[0007] This invention aims to provide a security assessment method for unmanned aerial vehicle (UAV) swarm systems based on multi-layer complex network theory, addressing the challenges of low-altitude airspace complexity and risks brought about by the rapid development of UAV technology. With the widespread application of UAVs in the civilian sector, the diversification of their missions and the increased complexity of low-altitude airspace have led to higher dynamic evolution characteristics and the possibility of cross-domain risk propagation in UAV swarm systems. In this context, traditional security assessment methods are no longer sufficient to meet the needs for accurate and comprehensive evaluation of UAV swarm systems.

[0008] This invention constructs a multi-layered complex network model, abstracting the task layer and structural layer of a UAV swarm, and considering its dynamic characteristics, various failure modes, and task dependencies. This allows for a more accurate description of the complex structure and interactions of UAV swarm systems. Through an improved dynamic propagation model, the invention simulates the propagation and recovery processes of faults within the system, quantitatively assesses the system's safety performance, and explores the quantitative relationship between system safety and propagation parameters. This invention not only accurately assesses the safety of UAV swarm systems but also provides a scientific basis for optimized system management, fault recovery, and informed decision-making. It improves the operational efficiency and reliability of UAV swarms in complex low-altitude scenarios, strongly supports the widespread application of UAV technology, and ensures that UAVs reliably provide services to society.

[0009] (II) Technical Solution

[0010] This invention provides a security assessment method for unmanned aerial vehicle (UAV) swarm systems based on multi-layer complex network theory, the steps of which are as follows:

[0011] Step 1: Based on the characteristics of UAV swarms, abstract the UAV mission and UAV structure into complex network models, analyze the topological characteristics of the network, and define the fault types of single-layer networks;

[0012] Step 2: Use a traditional dynamic propagation model to simulate the recovery process of the UAV mission layer and structural layer network after being disturbed, and explore the dynamic propagation mechanism of the fault;

[0013] Step 3: Analyze the fault propagation process and explore the relationship between the critical point and propagation parameters of the UAV swarm system based on the traditional dynamic propagation model for individual task layer networks and structural layer networks;

[0014] Step 4: Combining the interdependent task layer and structure layer networks, and considering the directionality of links based on multi-layer network theory, construct a mutually coupled two-layer network for UAV swarms, supplement the fault types of the two-layer network, and explore the network structure characteristics.

[0015] Step 5: Improve the propagation model, analyze the propagation process of faults within and between layers, and evaluate the two-layer drone swarm.

[0016] The specific implementation of "based on the characteristics of UAV swarms, abstracting UAV tasks and UAV structures into complex network models, analyzing the topological characteristics of the network, and defining single-layer network fault types" in step 1 is as follows: UAV swarms can be abstracted into a task layer and a structure layer. The task layer is responsible for task execution, such as aerial rescue and water spraying, while the structure layer represents the structure of the UAV swarm.

[0017] The data structure of a network is typically described by an adjacency matrix in graph theory; that is, the constructed task-layer network is represented as G. a=(V a E a The structural layer network is represented as G. b =(V b E b ), where V a V b Record the node information in the corresponding network, E a E b Record the link relationships between corresponding network nodes. Assume there are n drones in the drone swarm, performing m types of tasks. Each drone can only perform one task (carrying one payload). Therefore, the number of nodes in both the task layer and the structure layer is n. The node information in the drone's task layer network is stored in a list. Node information in the UAV structural layer network is stored in a list.

[0018] Considering the actual network characteristics, for the task-layer network, nodes represent the workload of performing tasks (such as radar, cameras, sensors, weapons, etc.). Nodes performing the same task form a fully connected network. Based on the task level, nodes are divided into high-level nodes and low-level nodes. For the structure-layer network, considering the dynamic characteristics of UAV swarms, the distance between UAVs changes over time. An activity-driven model is used to model the dynamic characteristics of the UAV swarm as a dynamic network, where nodes represent UAVs and edges represent the distances between UAVs. For both the task-layer and structure-layer networks, the link relationships between nodes are represented by adjacency matrices A[n][n] and B[n][n], respectively, composed of two-dimensional arrays. For the task-layer network, if there is a link relationship between node i and node j, then the elements in adjacency matrix A[n][n] are... Nodes i and j are each other's adjacent nodes; they are neighbors. If no link exists, then... The relationship is represented as follows:

[0019]

[0020] For a structural layer network, if there is a link between node i and node j, then the elements in the adjacency matrix B[n][n] are... If there is no link relationship, then The relationship is represented as follows:

[0021]

[0022] Then, structural analysis was performed on the abstracted task layer network and structure layer network topology, and performance indicators, such as the degree of nodes, were calculated to characterize the structure of the UAV's task layer network and structure layer network.

[0023] For the task layer network of a drone swarm, only internal fault P1 is considered, which is the failure of the payload (such as radar, camera, sensor, weapon, etc.) that performs the task.

[0024] For a drone swarm structured layer network, two types of failures may occur. One is an internal failure, where the drone itself malfunctions, with a probability of P². The other is an intra-layer failure, with a probability of P. w If many other drones in neighboring nodes malfunction, then this drone is also highly likely to malfunction, considering factors such as weather (e.g., thunderstorms, typhoons) and terrain (complex flight environments, prone to collisions with obstacles). Considering the characteristics of activity-driven structure layer networks, P... w The corresponding formula is as follows:

[0025]

[0026] Wherein, the baseline infection rate (I0) represents the initial probability of infection, and the interaction intensity (I... inter ): A coefficient used to scale based on interaction weights, edge weights (w) ij ): Represents the strength or frequency of the relationship between nodes i and j in the network, with maximum infection adjustment (I max ): The upper limit of the infection rate, used to normalize I0 and I0. ihter The additional adjustment factor (η) may represent external factors that affect the probability of infection, such as environmental conditions or time effects.

[0027] The specific implementation of step 2, "simulating the recovery process of the UAV mission layer and structural layer networks after interference using a traditional dynamic propagation model to explore the dynamic propagation mechanism of faults," is as follows: Consider using the traditional dynamic propagation model SIR (Susceptible-Infected-Recovered) to simulate fault propagation. During the evolution process, the mission layer and structural layer networks have nodes in three states: susceptible nodes (S), faulty nodes (I), or recoverable nodes (R). Susceptible nodes (S) are normally functioning nodes but may fail due to the influence of adjacent faulty nodes; faulty nodes (I) are already failed nodes that can propagate faults to adjacent susceptible nodes with a certain probability; recoverable nodes (R) recover from faults, possess immunity, and are no longer affected by fault propagation. In the traditional dynamic propagation SIR model, the state transition rules include two parts. One is fault propagation, where faulty nodes (I) can propagate faults with probability β to adjacent susceptible nodes (S), corresponding to the process of some nodes in the mission layer and structural layer networks changing from the susceptible (S) state to the faulty (I) state. The second is fault recovery. A faulty node (I) can recover to a normal working (R) state with probability μ and will not fail again. This corresponds to the process in which some nodes in the task layer network and the structure layer network change from the faulty (I) state to the recovered (R) state. Since the state of the recovered (R) nodes does not change, the evolution process can be described as S→I→R. Eventually, all nodes in the task layer network and the structure layer network become recovered (R), and the total number of nodes in each network remains n.

[0028] Specifically, step 3, "analyzing the fault propagation process and exploring the relationship between the critical point and propagation parameters of the UAV swarm system based on the traditional dynamic propagation model for individual task-level and structural-level networks," is implemented as follows: Based on the traditional propagation model, the dynamic mechanism of faults is explored for both the task-level and structural-level networks. Applying the traditional dynamic propagation model from step 2, vulnerable individuals (S) in the UAV swarm system are transformed into faulty individuals (I), infecting other systems. Simultaneously, the faulty individuals (I) transform into recoverable individuals (R), and their states no longer change. This step primarily explores the recovery process of the UAV swarm system as it begins to "resist disturbances," setting the critical duration T... C As a security evaluation metric for the system, for task-layer networks, T = T C1 When the total number of faulty nodes in the task layer network is 0, it is considered to have reached a critical state, at which point the performance of the UAV swarm system has fully recovered to normal operation; for the structure layer network, T = T C2 When the total number of faulty nodes in the structural layer network is 0, it is considered to have reached a critical state. At this time, the performance of the UAV swarm system has fully recovered to normal operation. The critical time T is then investigated. C1 and T C2The relationship with the recovery rate μ.

[0029] Specifically, the process described in step 4, "combining the interdependent task layer and structure layer networks, considering the directionality of links based on multi-level network theory, constructing a mutually coupled two-layer UAV swarm network, supplementing the fault types of the two-layer network, and exploring the network structure characteristics," is as follows: Step 1 establishes two networks for the UAV swarm: a task layer network and a structure layer network. Based on multi-level complex network theory, considering the directionality of links and the mutual coupling relationship, a two-layer UAV swarm network is established. The links between layers represent the relationship between the UAVs and the corresponding task execution. The constructed mutually coupled two-layer UAV swarm network is represented as G = {G...} a G b Furthermore, there exists a matrix representing the coupling relationship, called the interlayer adjacency matrix C, where E a,b The edges representing the nodes of the task layer and the structure layer are represented in C as follows:

[0030]

[0031] For networks abstracted from various drone swarms, whether the links are directional or have weights, the above method can be used for network representation. Specifically, the link between node i and node j has a weight W. ij Then the corresponding adjacency matrix is ​​updated as follows: Regarding the directionality of links, a multi-layered UAV swarm network can be constructed that considers different link types, such as intra-layer directed (Directed) and inter-layer undirected (Undirected) and inter-layer directed (Directed) links.

[0032] Combining the task layer network and structure layer network established in step 1, the constructed UAV swarm two-layer network adopts an undirected topology within layers and undirected between layers, with 2^n nodes, represented as G = (V, E), and node information is stored in a list. In the task layer and structure layer networks, intra-layer links maintain the topology from step 1. Considering that structure layer nodes represent UAV entities and task layer nodes represent task payloads, and each UAV has only one task payload, a one-to-one inter-layer link is established between the UAV and its carried task payload. Corresponding to its carrying load And so on, correspond In the adjacency matrix C of a two-layer UAV swarm network, the elements between layers a and b are represented as follows:

[0033]

[0034] This rigorous two-layer coupled network architecture maintains the independence of the physical and logical layers while providing a precise modeling foundation for subsequent cross-layer fault propagation analysis through an explicit "drone-load" mapping relationship. Based on this abstract layering, a corresponding two-layer network for drone swarms is constructed, and characteristic metrics representing multi-level networks, such as average node degree, are explored.

[0035] Building upon this, an additional type of fault may occur in a two-layer UAV swarm network: inter-layer faults. Faults in the structural layer network upon which task-layer nodes depend are related to faults in the task-layer nodes themselves, specifically those with the parameter P. mis The probability of failure occurs when the task layer network on which the structural layer node depends fails, and the corresponding structural layer node is represented by P. str The probability of failure occurs.

[0036] The specific steps for "improving the propagation model, analyzing the fault propagation process within and between layers, and evaluating the two-layer UAV swarm" described in step 5 are as follows: The traditional dynamic propagation model SIR is extended to a multi-layer network cascaded dynamic propagation model SIRD (Susceptible-Infected-Recovered-Delete). Improvements are made in two main aspects: First, unlike the traditional SIR model which only changes node states, the improved cascaded dynamic propagation model adds a new deletion state (D) to account for UAV crashes and forced landings, thus removing the corresponding node from the two-layer UAV swarm network and better reflecting real-world scenarios. Second, considering task levels, different state transition strategies are formulated for the task layer network and the structural layer network. For the task layer network, if a high-level node… Inter-layer faults occur, meaning they are affected by the corresponding nodes in the structural layer network. If the impact causes a malfunction, immediately use P. high The probability of a node becoming a vulnerable individual (S) is considered. If the transformation fails, the corresponding structural layer and task layer nodes are deleted. If a low-level node experiences an inter-layer failure, the corresponding structural layer and task layer nodes are deleted. For structural layer networks, if a high-level node... An inter-layer fault occurs, meaning it is affected by the corresponding node in the task layer network. If the impact causes a malfunction, then immediately use I high The probability of it recovering to a normal (R) state and not failing again is high; if recovery fails, the corresponding structural layer and task layer nodes are deleted; if a low-level node experiences an inter-layer failure, it is immediately replaced by an I... low The probability of becoming a vulnerable individual (S) changes, and if the transformation fails, the corresponding structural layer and task layer node is deleted. In this case, the evolutionary process becomes... Ultimately, the number of nodes in the two-layer network of the drone swarm that become either in a recovered or deleted state is less than or equal to 2n.

[0037] Similarly, the critical duration TC As a security assessment indicator for the entire drone swarm system, for a two-layer drone swarm network, T = T C When the total number of faulty nodes in the entire network is 0, it is considered to have reached a critical state, at which point the system performance has fully recovered to normal operation. The critical time T is then investigated. C The relationship with the recovery rate μ.

[0038] This invention calculates the critical time T required for a node in a two-layer UAV swarm network to recover from a failed state to full recovery to normal operation during the recovery process. C This allows for a quantitative assessment of the system's security performance, at which point the multi-layered drone swarm network has no fault-free nodes.

[0039] Through the above steps, a security assessment method for UAV swarms based on multi-layer complex network theory can be constructed. This invention, based on multi-layer complex network theory, abstracts the research object into a complex network layer by layer. By improving the dynamic propagation model, it simulates the recovery process of nodes in the system returning to normal operation after fault propagation, explores the fault propagation mechanism, and establishes a critical time T when the proportion of faulty nodes in the system is exactly 0. C A quantitative assessment of the security performance of unmanned aerial vehicle (UAV) swarm systems was conducted.

[0040] (III) Advantages and Efficacy of the Invention

[0041] The advantages of this invention compared to existing technologies are as follows: This invention considers the various failure modes of UAV swarms and the dynamic characteristics of UAV structures, while distinguishing UAVs of different mission levels. Based on a research paradigm of multi-layer complex network theory from a systems perspective, it quantitatively evaluates the safety performance of UAV swarm systems. Previous studies based on complex network theory for UAV swarm safety assessments could not effectively quantitatively assess the safety performance of UAV structures and missions in withstanding potential dangers, absorbing damage, and recovering to normal operation. They lacked consideration of the dynamic characteristics of UAV swarms and a hierarchical UAV system safety performance assessment framework. To address these issues, this invention combines complex network theory to abstractly model single-layer and multi-layer coupled UAV swarms. For the mission-layer network, UAVs of different mission levels execute different fault recovery strategies; considering this characteristic, it is modeled as a static network that distinguishes UAV levels. For the structural-layer network, considering that the distance between UAVs in the swarm changes, it is modeled as a dynamic network. The multi-layer complex network theory breaks through the limitation of homogeneity of nodes and links in single-layer networks, and can more accurately describe UAV swarm systems in real society that contain multiple types (intra-layer and inter-layer) of link relationships. After constructing the corresponding network model, this invention simulates the system recovery process under the influence of faults by improving the dynamic propagation model, calculates the critical time for the UAV swarm system to resume normal operation, and quantitatively evaluates the system's safety performance. This leads to the construction of a general safety assessment framework for single-mode and multi-level UAV swarm systems. This invention provides a basis for effectively assessing and improving the operational capabilities of UAV swarm systems. Furthermore, this framework provides decision-makers with a new approach and guiding direction for designing optimized management schemes for UAV swarm systems, thereby further ensuring that UAV swarm systems provide services in a reliable manner. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the method flow described in this invention.

[0043] Figure 2A This is a visual representation of the task layer network abstractly constructed by the UAV swarm system in this embodiment of the invention.

[0044] Figure 2B This is a visual representation of the structural layer network abstractly constructed by the UAV swarm system in this embodiment of the invention.

[0045] Figure 3A The critical time T of the UAV cluster task layer network in this embodiment of the invention. C1 A visual representation of the relationship between the recovery rate μ and the recovery rate.

[0046] Figure 3B The critical time T of the UAV swarm structure layer network in this embodiment of the invention.C2 A visual representation of the relationship between the recovery rate μ and the recovery rate.

[0047] Figure 4 This is a visual representation of the two-layer complex network abstractly constructed by the UAV swarm system in this embodiment of the invention.

[0048] Figure 5 The critical time T of the two-layer network for the UAV swarm in this embodiment of the invention. C A visual representation of the relationship between the recovery rate μ and the recovery rate. Detailed Implementation

[0049] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0050] This invention provides a security assessment method for UAV swarm systems based on multi-layer complex network theory. It abstracts and models the interdependencies between UAV swarm tasks and structural networks using nodes and links, simulates the dynamic propagation of hazards such as faults by improving the dynamic propagation model, and quantitatively assesses the security performance of UAV swarm systems recovering from a failure state to a normal working state under the influence of hazards. The constructed general security assessment framework helps to optimize the management of multi-layer and multi-mode UAV swarm systems and improve operational reliability.

[0051] like Figure 1 As shown, the present invention provides a method for security assessment of a drone swarm system based on multi-layer complex network theory, comprising the following steps:

[0052] Step 1: Based on the characteristics of UAV swarms, abstract the UAV mission and UAV structure into complex network models, analyze the network topology characteristics, and define the fault types of single-layer networks.

[0053] The relative positions of drones in a drone swarm can change as tasks are executed. Considering its dynamic characteristics, and the diverse tasks performed by drones (different models can perform different tasks), a drone swarm can be abstracted into two networks: a task layer and a structure layer. The task layer is responsible for task execution, such as aerial rescue and water spraying, while the structure layer corresponds to the dynamic structure of the drone swarm. In modeling the task layer network, considering that different task levels correspond to different fault propagation strategies, nodes are divided into high-level and low-level nodes based on their task level. Figure 2AAs shown in the diagram. During the simulation, the task layer network has 200 nodes. Ten fully connected networks are established for ten different tasks, with five corresponding to high-level tasks and five to low-level tasks. For the structure layer network, considering the dynamic characteristics of the UAV swarm, the distance between UAVs changes over time. An activity-driven model is used to model the dynamic characteristics of the UAV swarm as a dynamic network. The structure layer network changes continuously at each moment, as shown in the diagram. Figure 2B As shown, the simulation involved a total of 200 nodes. This invention is based on a constructed single-layer task layer and structure layer network model, analyzes the corresponding topological characteristics, and characterizes these topological characteristics using metrics such as node degree and average degree.

[0054] ● The degree of the node is Average degree

[0055] For the task layer network of a drone swarm, only internal faults P1=0.1 are considered.

[0056] Consider two types of faults in a drone swarm architecture layer network. The probability of an internal fault is P² = 0.1. The probability of a layer-specific fault is P... w The corresponding formula parameter is set to I0 = 0.1, I inter =0.05, I max =3, η=1, where the edge weight w ij The weights of the edges at each time step in the activity-driven network.

[0057]

[0058] Step 2: Use a traditional dynamic propagation model to simulate the recovery process of the UAV mission layer and structural layer network after being disturbed, and explore the dynamic propagation mechanism of the fault.

[0059] Based on a complex network model abstracted from the structure layer and task layer of a drone swarm, this invention considers the traditional SIR (Susceptible-Infected-Recovered) dynamic propagation model to simulate fault propagation and the recovery process of a drone swarm system after node failure due to faults or other hazards. In this invention, for the structure layer network, fault propagation within the layer is considered. If more than half of the drones in a neighboring node fail, then that drone is also highly likely to fail. The failed node (I) can transmit the fault with probability β = P. w The fault propagation occurs to adjacent nodes of the vulnerable node (S). Simultaneously, the failed node (I) can recover to its normal operating state (R) with probability μ = 0.2 and will not fail again. The fault propagation process is as follows:

[0060] In the initial phase, it is assumed that 10% of the nodes are in a failed state, and 90% of the nodes are in a normal operating state. Subsequently, the task layer and structural layer networks are evolved according to the corresponding failure modes. It is important to note that for the task layer network, since intra-layer failures are not considered, its failures are self-inflicted, rather than propagated from failed nodes (I).

[0061] Step 3: Analyze the fault propagation process and explore the relationship between the critical point and propagation parameters of the UAV swarm system based on the traditional dynamic propagation model for individual task layer networks and structural layer networks.

[0062] By simulating the fault propagation process, this invention calculates the impact of fault recovery rate on node recovery time in both the task layer and structural layer networks of an unmanned aerial vehicle (UAV) swarm. Furthermore, it calculates the critical time T for a node in an UAV swarm system to fully recover from a failed state to normal operation. C To quantitatively measure the security performance of a drone swarm system, in T=T C At this point, the failed / faulty nodes in the UAV swarm system have just fully recovered, meaning the proportion of failed nodes is 0. The critical time for the task-layer network is denoted as T. C1 The critical time of the structural layer network is denoted as T. C2 The simulation experiment simulated the recovery process of an unmanned aerial vehicle (UAV) swarm system after a failure and explored the critical time T. C1 and T C2 The relationship with the recovery rate μ is as follows: Figure 3A and Figure 3B As shown.

[0063] Step 4: Combining the interdependent task layer and structure layer networks, and considering the directionality of links based on multi-layer network theory, construct a coupled two-layer network for UAV swarms, supplement the fault types of the two-layer network, and explore the network structure characteristics.

[0064] This invention, based on multi-layer network theory, considers the directionality of links and mutual coupling relationships to establish a two-layer network for UAV swarms. In reality, considering that in UAV swarm systems, the task layer and the structure layer are interdependent, task layer nodes require dynamic changes in structure layer nodes to execute different tasks, while the structure layer network relies on the workload of the task layer network, such as GPS, for positioning to achieve collision avoidance. Based on this, an interdependent two-layer network for UAV swarms is constructed, with a one-to-one correspondence between structure layer nodes and task layer nodes. The links between layers represent the interdependence between the task layer and the structure layer, such as... Figure 4 As shown, there are 400 nodes in the simulation. The constructed mutually coupled two-layer UAV swarm network is represented as G = {G...} a G b}, inter-layer adjacency matrix C, the links within and between layers in the UAV swarm system in this invention example are all undirected.

[0065] In a two-layer network of a drone swarm, inter-layer faults are considered as follows: A fault in the structural layer network upon which a task layer node depends is addressed by a fault in the corresponding task layer node (e.g., P). mis A failure occurs with a probability of 0.8. A failure in the task layer network upon which a structural layer node depends corresponds to a structural layer node with a probability of P. str The probability of failure is 0.2.

[0066] Step 5: Improve the propagation model, analyze the propagation process of faults within and between layers, and evaluate the two-layer drone swarm.

[0067] Based on a multi-layered coupled UAV swarm network, the traditional dynamic propagation model SIR in step 2 is extended to a multi-layered network cascaded dynamic propagation model SIRD with the addition of a new state D. The inter-layer fault modes of the task layer and the structural layer are extended respectively. The specific implementation is as follows.

[0068] In the initial stage, 10% of the nodes in the task layer and structure layer network of the UAV swarm are randomly selected and set to a failed state, while the remaining 90% of the nodes are in a normal working state. The node recovery rate is set to μ = 0.2.

[0069] Consider the fault propagation process within and between layers. In addition to the faults considered in step 2, inter-layer faults also need to be considered, evolving differently according to the task level. For a task-level network, if an inter-layer fault occurs in a high-level node, it will immediately propagate via P... high A node with a probability of 0.8 becomes a vulnerable individual (S). If it fails, the corresponding structural and task layer nodes are deleted. If a lower-level node experiences an inter-layer failure, the corresponding structural and task layer nodes are also deleted. For structural layer networks, if a higher-level node experiences an inter-layer failure, it is immediately replaced by an I-level node. high With a probability of 0.8, it recovers to a normal (R) state and will not fail again. If recovery fails, the corresponding structural layer and task layer nodes are deleted. If a low-level node experiences an inter-layer failure, it immediately uses I... low If the probability is 0.2, it becomes a vulnerable individual (S). If it fails, the corresponding structural layer and task layer node is deleted.

[0070] Simultaneously, during the node recovery process, the failed node (I) recovers to a normal operating (R) state with a probability of μ = 0.2 and will not fail again. Similar to step 3, the impact of different propagation rates on the security of the UAV swarm system is first explored based on the intra-layer and inter-layer propagation rates. This invention calculates the time required for a node in the two-layer network of the UAV swarm to fully recover from a failed state to a normal operating state, i.e., the critical time T. C This allows for a quantitative assessment of the security performance of drone swarm systems.

[0071] Through the above steps, a security assessment method for UAV swarms based on multi-layer complex network theory can be constructed. This invention, based on multi-layer complex network theory, abstracts the research object into structural and task layers. By improving the dynamic propagation model, it simulates the recovery process of nodes in a UAV swarm system after fault propagation to normal operation, exploring the fault propagation mechanism, and determining the critical time T when the proportion of faulty nodes in the UAV swarm system is exactly 0. C A quantitative assessment of the security performance of unmanned aerial vehicle (UAV) swarm systems was conducted.

[0072] This invention defines the critical time T for restoring a drone swarm system to normal operation. C As a security assessment indicator for unmanned aerial vehicle (UAV) swarm systems, and to explore the critical time T C The relationship between the recovery rate μ and the recovery rate, such as Figure 5 As shown, G a Corresponding to the task layer, G b In the corresponding structural layer, G corresponds to a two-layer network. high Corresponding to the high-level network of the task layer, G low This corresponds to a lower-level network at the task layer. It can be seen that the larger the value of μ, the smaller the critical time, and the shorter the time required for the drone swarm system to recover to normal operation. T C When both μ and μ are taken as logarithms, they show a linear relationship. At the same time, it can be observed that the critical time of high-level tasks is significantly shorter than that of low-level tasks, which ensures the smooth execution of high-level tasks.

[0073] Through the above steps, a security assessment method for UAV swarm systems based on complex network theory can be constructed. This technology fully considers the dynamic characteristics of UAV swarms and the differences in mission levels, and uses multi-layer complex network theory to abstractly model the UAV swarm system. Through an improved dynamic propagation model, the propagation process of faults within and between layers in the two-layer network of the UAV swarm is simulated, quantitatively evaluating the security performance of UAV swarm system components recovering from a failed state to a normal operating state. The general security assessment framework constructed in this invention is applicable to different types of UAV swarm systems, helps optimize system management, improves operational reliability, and provides new ideas and methods for developing effective fault management solutions.

[0074] The parts of this invention not described in detail are well-known in the field.

[0075] The above description is only a part of the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A security assessment method for unmanned aerial vehicle (UAV) swarm systems based on multi-layer complex network theory, characterized in that, The steps include the following: Step 1: Based on the characteristics of UAV swarms, abstract the UAV mission and UAV structure into complex network models, analyze the topological characteristics of the network, and define the fault types of single-layer networks; Step 2: Use a traditional dynamic propagation model to simulate the recovery process of the UAV mission layer and structural layer network after being disturbed, and explore the dynamic propagation mechanism of the fault; Step 3: Analyze the fault propagation process and explore the relationship between the critical point and propagation parameters of the UAV swarm system based on the traditional dynamic propagation model for individual task layer networks and structural layer networks; Step 4: Combining the interdependent task layer and structure layer networks, and considering the directionality of links based on multi-layer network theory, construct a mutually coupled two-layer network for UAV swarms, supplement the fault types of the two-layer network, and explore the network structure characteristics. Step 5: Improve the propagation model, analyze the propagation process of faults within and between layers, and evaluate the two-layer drone swarm.

2. The method for security assessment of a UAV swarm system based on multi-layer complex network theory according to claim 1, characterized in that: In step 1, the drone swarm is abstracted into a task layer and a structure layer. The task layer is responsible for task execution, and the structure layer represents the structure of the drone swarm. The network's data structure is described by the adjacency matrix in graph theory; that is, the constructed task-layer network is represented as G. a =(V a E a The structural layer network is represented as G. b =(V b E b ), where V a V b Record the node information in the corresponding network, E a E b Record the link relationships between corresponding network nodes; assuming there are n drones in the drone swarm, performing m types of tasks, and each drone can only perform one task (carrying one payload), then the number of nodes in both the task layer and the structure layer is n. The node information in the drone's task layer network is then stored in a list. Node information in the UAV structural layer network is stored in a list.

3. The security assessment method for a drone swarm system based on multi-layer complex network theory according to claim 1, characterized in that: For a task-layer network, nodes represent the load of executing tasks. Nodes executing the same task form a fully connected network. Based on the task level, nodes are divided into high-level nodes and low-level nodes. For the structural layer network, considering the dynamic characteristics of the UAV swarm, the distance between UAVs changes over time. An activity-driven model is used to model the dynamic characteristics of the UAV swarm as a dynamic network, where nodes represent UAVs and edges represent the distance between UAVs. For the task layer network and the structural layer network, the connection relationship of nodes is represented by the adjacency matrices A[n][n] and B[n][n] composed of two-dimensional arrays, respectively. For a task-layer network, if there is a link between node i and node j, then the elements in the adjacency matrix A[n][n] are... Nodes i and j are each other's adjacent nodes; they are neighbors. If no link exists, then... The relationship is represented as follows: For a structural layer network, if there is a link between node i and node j, then the elements in the adjacency matrix B[n][n] are... If there is no link relationship, then The relationship is represented as follows: Then, structural analysis was performed on the abstracted task layer network and structure layer network topology, and performance indicators characterizing the structure of the UAV's task layer network and structure layer network were calculated.

4. A method for security assessment of a drone swarm system based on multi-layer complex network theory according to claim 1, 2, or 3, characterized in that: For the task layer network of a drone swarm, only internal fault P1 is considered, that is, the load executing the task fails. For a drone swarm architecture layer network, two types of failures may occur: one is an internal failure, i.e., a failure of the drone itself, with a probability of P2; the other is an intra-layer failure, with a probability of P. w If many other drones in a neighboring node fail, then this drone is also very likely to fail. Considering the characteristics of activity-driven structure layer networks, P w The corresponding formula is as follows: Wherein, the baseline infection rate I0 represents the initial probability of infection, and the interaction intensity I... inter A coefficient is used to scale based on interaction weights, where w is the edge weight. ij This represents the strength or frequency of the relationship between nodes i and j in the network, with the maximum infection adjustment I. max This represents the upper limit of the infection rate, used to normalize I0 and I... inter The additional adjustment factor η represents the external factors affecting the probability of infection.

5. The method for security assessment of a drone swarm system based on multi-layer complex network theory according to claim 1, characterized in that: In step 2, the traditional dynamic propagation model SIR is used to simulate fault propagation. During the evolution process, the task layer network and the structural layer network have nodes in three states: vulnerable nodes S, faulty nodes I, or recoverable nodes R. Failure-prone node S is a normally functioning node, but may fail due to the influence of adjacent faulty nodes; Faulty node I is a failed node that can propagate the fault to adjacent failure-prone nodes; Recovery node R recovers from the fault, is immune, and is no longer affected by the propagation of the fault.

6. The method for security assessment of a drone swarm system based on multi-layer complex network theory according to claim 5, characterized in that: In the SIR model, the state transition rule consists of two parts: first, fault propagation, where faulty node I propagates the fault to its neighboring nodes in the vulnerable state S with probability β, corresponding to the process where some nodes in the task layer network and the structure layer network change from the vulnerable state S to the faulty state I; second, fault recovery, where faulty node I recovers to the normal working state R with probability μ and will not fail again, corresponding to the process where some nodes in the task layer network and the structure layer network change from the faulty state I to the recovered state R. Since the state of the recovered R node does not change, the evolution process is described as S→I→R, and eventually all nodes in the task layer network and the structure layer network become the recovered state R, and the total number of nodes in each network remains n.

7. A method for security assessment of a drone swarm system based on multi-layer complex network theory as described in claim 5 or 6, characterized in that: In step 3, the vulnerable individual S in the drone swarm system is transformed into a faulty individual I and infects the drone. At the same time, the faulty individual I is transformed into a recoverable individual R and its state will not change again. The critical duration T C As a security evaluation metric for the system, for task-layer networks, T = T C1 When the total number of faulty nodes in the task layer network is 0, it is considered to have reached a critical state, at which point the performance of the UAV swarm system has fully recovered to normal operation; for the structure layer network, T = T C2 When the total number of faulty nodes in the structural layer network is 0, it is considered to have reached a critical state. At this time, the performance of the UAV swarm system has fully recovered to normal operation. The critical time T is then investigated. C1 and T C2 The relationship with the recovery rate μ.

8. The method for security assessment of a drone swarm system based on multi-layer complex network theory according to claim 1, characterized in that: In step 4, the constructed mutually coupled two-layer UAV swarm network is represented as G = {G} a G b Furthermore, there exists a matrix representing the coupling relationship, called the interlayer adjacency matrix C, where E a,b The edges representing the nodes of the task layer and the structure layer are represented in C as follows: Among them, the link between node i and node j has a weight W. ij Then the corresponding adjacency matrix is ​​updated as follows: Regarding the directionality of links, a multi-layered UAV swarm network is constructed that considers different link types, including intra-layer directed and inter-layer undirected, and intra-layer undirected and inter-layer directed.

9. The method for security assessment of a UAV swarm system based on multi-layer complex network theory according to claim 8, characterized in that: The constructed two-layer UAV swarm network adopts an undirected topology within each layer and undirected between layers, with 2^n nodes, represented as G = (V, E). Node information is stored in a list. In the middle; for the task layer and structure layer networks, the intra-layer links maintain the topology in step 1; considering that the structure layer nodes represent UAV entities and the task layer nodes represent task payloads, and each UAV has only one task payload, a one-to-one inter-layer link is established between the UAV and its carried task payload, and the UAV node Corresponding to its carrying load And so on, correspond In the adjacency matrix C of a two-layer UAV swarm network, the elements between layers a and b are represented as follows: The two-layer network of UAV swarms constructed based on this introduces a new type of fault: inter-layer faults, which are faults in the structural layer network upon which task-layer nodes depend. These faults correspond to faults in the task-layer nodes with the parameter P. mis The probability of failure occurs when the task layer network on which the structural layer node depends fails, and the corresponding structural layer node is represented by P. str The probability of failure occurs.

10. The method for security assessment of a drone swarm system based on multi-layer complex network theory according to claim 1, characterized in that: In step 5, the traditional dynamic propagation model SIR is extended to the multi-layer network cascaded dynamic propagation model SIRD. The cascaded dynamic propagation model adds a new deletion state D, which is to delete the corresponding node from the two-layer network of the UAV swarm. At the same time, different state transition strategies are formulated for the task layer network and the structure layer network respectively. For task-level networks, if high-level nodes Inter-layer faults occur in i∈[1,n], i.e., they are affected by the corresponding nodes of the structural layer network. If the impact causes a malfunction, then immediately use P. high If the probability of a node becoming a vulnerable individual S is lost, and the transformation fails, the corresponding structural layer and task layer nodes are deleted. If a low-level node experiences an inter-layer failure, the corresponding structural layer and task layer nodes are deleted. For structural layer networks, if a high-level node... Inter-layer faults occur in i∈[1,n], i.e., they are affected by the corresponding nodes in the task layer network. If the impact causes a malfunction, then immediately use I high The probability of it recovering to the normal working R state and not failing again is high; if recovery fails, the corresponding structural layer and task layer nodes are deleted; if a low-level node experiences an inter-layer failure, it is immediately replaced by I. low The probability of becoming a vulnerable individual S is changed; if the transformation fails, the corresponding structural layer and task layer nodes are deleted. The evolutionary process becomes Ultimately, the number of nodes in the final two-layer drone swarm network that either return to a recovered or deleted state is less than or equal to 2n; the critical duration T is then set. C As a security assessment indicator for the entire drone swarm system, for a two-layer drone swarm network, T = T C When the total number of faulty nodes in the entire network is 0, it is considered to have reached a critical state, at which point the system performance has fully recovered to normal operation. The critical time T is then investigated. C The relationship with the recovery rate μ.

Images