Unmanned aerial vehicle cluster network damage evaluation method based on entropy weight method-TOPSIS

Abstract

This invention discloses a damage assessment method for UAV swarm networks based on the entropy weight method-TOPSIS. For UAV swarm networks and for damaged nodes, it obtains secondary indicator values ​​under four primary indicators: resilience, real-time performance, survivability, and communication recovery capability, constructing an indicator matrix corresponding to each primary indicator. Each primary indicator matrix is ​​standardized, and then weighted using entropy-related weight vectors to obtain an entropy-weighted standardized matrix for each primary indicator. The value score for each primary indicator is then calculated, and a comprehensive score for the damaged node is obtained by weighting the entropy-related weight vectors. The scores of each node's neighboring nodes are recursively accumulated, and this, combined with the comprehensive score of the currently damaged node, yields the final score for the damaged node. The final scores of all damaged nodes serve as the damage assessment result for the UAV swarm network.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm network technology, and more specifically to a damage assessment method for UAV swarm networks based on the entropy weight method-TOPSIS. Background Technology

[0002] In recent years, drone swarm warfare technology has become a key technology for achieving precision strikes in modern warfare. When facing drone swarm attacks, the defending side needs to scientifically and accurately simulate the damage effects on the target, conduct multiple assessments of the damage caused by different attack methods, identify the best attack timing and targets, scientifically arrange the strike plan, and formulate a clear and specific combat plan.

[0003] To accurately perceive the damage effects of drones, many predecessors have proposed various evaluation metrics when studying drone swarm networks. Some works are based on basic metrics of ordinary complex networks, while others have proposed and constructed entirely new network metrics to evaluate specific working environments and tasks. There are many existing studies and works on damage simulation of drone swarm networks, and the following are some of the more common ones:

[0004] To evaluate the applicability of existing routing protocols in UAV ad hoc networks, Dong Qiangjian et al. used OPNET network simulation software to construct three typical simulation scenarios and conducted routing protocol simulation analysis. The typical environments were divided into friendly, obstacle, and adversarial types, and five simulation scenarios were established. Performance parameters such as end-to-end latency, delivery rate, and network load were compared and analyzed. The results showed that different routing protocols can be applied to different UAV ad hoc network scenarios.

[0005] In the process of Li Cao et al. studying targeted routing protocols for swarm drone self-organizing networks, they adopted indicators such as packet loss rate, throughput, and average latency. While keeping other parameters unchanged, they changed different routing protocols to compare the performance of the drone swarm and observed the network status of the drone swarm by controlling variables.

[0006] Zhang Xin et al. proposed a medium-degree entropy resilience measurement method for evaluating complex networks. Based on node importance, they studied network resilience measurement and resilience performance, proposed a local medium-degree centrality index, and considered the node clustering coefficient to propose a node resilience measurement method. After damage experiments, they concluded that an attack strategy based on medium-degree centrality can approximately partition the network into a set of isolated nodes.

[0007] Wang Yuang et al. designed an infection-propagation model to address the multi-layered characteristics of UAV swarm networks. They also designed an efficiency index for single-layer complex networks using a path-based approach and used this index as the basis for multi-layer network analysis. They proposed a method for identifying important nodes in multi-layer complex networks based on layer weight-K-shell-Eigenvector centrality.

[0008] Li Yanjun et al. proposed a method for assessing the resilience of UAV swarms based on information interaction. Due to the inherent complexity of UAV swarm systems in battlefield combat environments, Tran's resilience assessment method based on complex networks only reflects the network's topology and is therefore unsuitable for UAV swarms. Li's proposed UAV swarm model based on information interaction presents a different approach. Building upon the complex network-based resilience assessment method, this method changes the comparison from assessing communication interaction capabilities to comparing the ideal state of swarm information interaction capabilities with the state before recovery and the stable state after recovery following the damage event.

[0009] The methods described above, when used for damage performance assessment of UAV swarm networks, lack a unified modeling system and are relatively simplistic, often relying solely on weighted PCA modeling or single-index scoring, lacking comprehensive consideration and, more importantly, hierarchical modeling. The Task Reliability (TOC) assessment method mentioned earlier, while considering the survivability and mission metrics of the UAV swarm itself, lacks consideration of the complex network communication capabilities. Existing simulation and assessment methods cannot accurately perceive the damage effects on UAVs. For damage simulation and assessment of UAV swarms, existing methods are not highly accurate and struggle to comprehensively evaluate the damage resistance of UAV swarm networks. Summary of the Invention

[0010] In view of this, the present invention provides a damage assessment method for UAV swarm networks based on the entropy weight method-TOPSIS. For damaged nodes in the UAV swarm network, the method obtains the values ​​of secondary indicators under four primary indicators: resilience, real-time performance, survivability, and communication recovery capability. From the four aspects of effectiveness, survivability, real-time performance, and communication recovery capability, a comprehensive score for each damaged node is obtained through the secondary entropy weight method-TOPSIS scoring, thereby achieving a comprehensive and high-precision evaluation of the damage resistance performance of the UAV swarm network.

[0011] To achieve the above objectives, the technical solution of this invention is as follows: a damage assessment method for UAV swarm networks based on the entropy weight method-TOPSIS. For UAV swarm networks, for damaged nodes, the values ​​of secondary indicators under four primary indicators are obtained: resilience, real-time performance, survivability, and communication recovery capability. For each primary indicator, an indicator matrix is ​​constructed with its corresponding secondary indicators as rows and each damaged node as a column, denoted as the primary indicator matrix.

[0012] Each primary indicator matrix is ​​standardized, and the standardized primary indicator matrix is ​​weighted using a weight vector related to the entropy value to obtain the entropy-weighted standardized matrix of the primary indicator.

[0013] The value score of each primary indicator is calculated based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicator.

[0014] For the value scores of each primary indicator, a comprehensive score for the damaged node is obtained by weighting the scores using a weight vector related to the entropy value.

[0015] A hop count threshold is preset. The current damaged node is searched for one-hop neighbor nodes. It is determined whether the hop count of the current neighbor node is less than the hop count threshold. If so, the sum of the comprehensive scores of the neighbor nodes with the current hop count is calculated and accumulated. The accumulated score is multiplied by a certain coefficient and then returned to the hop count threshold for judgment. When the preset hop count threshold is reached, the accumulated score is added to the comprehensive score of the current damaged node to obtain the final score of the damaged node.

[0016] The final score of all damaged nodes is used as the damage assessment result for the drone swarm network.

[0017] Furthermore, for the damaged nodes, the values ​​of the secondary indicators under the four primary indicators of resilience, real-time performance, survivability and communication assessment are obtained. Specifically, the primary indicator of resilience includes the following four secondary indicators: connectivity coefficient, average shortest path ratio, average clustering coefficient ratio and betweenness coefficient.

[0018] The primary metric, survivability, includes the following four secondary metrics: connectivity coefficient, end-to-end reliability, K-terminal reliability, and overall reliability.

[0019] The primary indicator of real-time performance includes the following four secondary indicators: throughput, transmission latency, packet loss rate, and average information transmission latency.

[0020] The primary indicator, communication recovery capability, includes the following secondary indicator: resilience.

[0021] Furthermore, the average clustering coefficient is C, specifically:

[0022] The number of damaged nodes in the cluster network is N, and the clustering coefficient of the i-th damaged node is C. i The number of neighboring nodes of the damaged node i is k. i The actual number of connections between neighbors is E. i Then the clustering coefficient C of node i i for:

[0023] Furthermore, the betweenness number is B. i The subscript i represents the i-th damaged node, i.e., B. i The betweenness of the damaged node i is specifically: Where g jk g is the number of shortest paths between damaged nodes j and k; jk (i) represents the number of shortest paths from j to k that pass through node i.

[0024] Furthermore, resilience specifically refers to:

[0025] Toughness S is

[0026] Among them, σ, δ, ρ, τ, Let [t0, t] represent the system's performance factor, absorption factor, recovery factor, recovery event factor, and volatility factor, respectively. final The communication performance of the UAV within a given time period is y(t) as the basis; the calculation methods for each factor are as follows:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032] y min The average number of messages received from the end of damage to the start of recovery; y D y represents the average number of messages received before the damage begins. R The number of messages received after recovery; t ss To restore the start time; SNR dB Signal-to-noise ratio (SNR) in network communication, measured in decibels (dB).

[0033]

[0034]

[0035]

[0036] t th At the moment of destruction, t min At the moment of destruction, t ss To resume from the start time.

[0037] Furthermore, each primary indicator matrix is ​​standardized, and the standardized primary indicator matrix is ​​weighted using a weight vector related to the entropy value to obtain the entropy-weighted standardized matrix of the primary indicators. This process includes the following steps:

[0038] S201: The element x in the i-th row and j-th column of each first-level indicator matrix. ij This represents the value of the i-th secondary indicator under the primary indicator at the j-th damage node, for each x. ij Based on the optimization direction of their corresponding secondary indicators, they are classified and unified into the same optimization direction to obtain positively processed indicator values. Among them, connectivity coefficient, average clustering coefficient ratio, end-to-end reliability, K-segment reliability, overall reliability, and throughput are all positive indicators, while average shortest path ratio, betweenness, transmission delay, packet loss rate, and average round-trip time are negative indicators. The negative indicators are then subjected to range positiveization processing.

[0039] S202: Standardize the index values ​​processed by positive transformation to obtain normalized index values;

[0040] S203: For the normalized index values, calculate the weight of each secondary index in the same secondary index at different damage nodes, and use the weight value to calculate the entropy value of the secondary index, so as to obtain the entropy value of each secondary index corresponding to each primary index in the four primary indexes of the UAV network damage model.

[0041] S204: Calculate the weights of the secondary indicators using the entropy values ​​of the secondary indicators. The weight combination of secondary indicators belonging to the same primary indicator is the weight vector of that primary indicator, which is the entropy-related weight vector. Use the entropy-related weight vector to weight the standardized primary indicator matrix to obtain the entropy-weighted standardized matrix of each primary indicator.

[0042] Furthermore, based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicators, the value score of each primary indicator is calculated. Specifically, based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicators, the Euclidean distance D between each primary indicator and the positive and negative ideal solutions is calculated. - and D + The value score of a node in the current primary indicator is D. - / (D - +D +).

[0043] Beneficial effects:

[0044] 1. This invention provides a method for assessing damage to UAV swarm networks based on the entropy weight method-TOPSIS. It takes data on various network indicators damaged at different nodes of the same network as an example. Using the four secondary indicators under the abstract indicator layer (throughput, transmission latency, packet loss rate, and average information transmission latency) as examples, the entropy weight method is applied. After forward standardization and normalization, the weight vector of each indicator is determined. Weight calculations are then performed to obtain the standard matrix for TOPSIS scoring. TOPSIS scoring is then performed to obtain the real-time score at the abstract layer. This is the first entropy weight method-TOPSIS scoring. After obtaining the scores for the four abstract layer indicators (real-time performance, survivability, resilience, and communication recovery capability), these four secondary indicators are used as metadata for a second entropy weight method-TOPSIS scoring. This involves normalizing these four indicators again to obtain the weight vector, and then performing TOPSIS scoring. The score obtained at this point is the comprehensive node score. Therefore, this invention achieves a comprehensive and high-precision evaluation of the damage resistance performance of UAV swarm networks by obtaining a comprehensive score for each damaged node from four aspects: effectiveness, survivability, real-time performance, and communication recovery capability through the secondary entropy weight method-TOPSIS scoring.

[0045] 2. Existing UAV swarm network damage assessment methods lack comprehensive network evaluation metrics. Most studies only select some or relevant metrics, with relatively simple modeling processes or direct use of general network evaluation metrics, lacking specificity and effectiveness. This invention establishes a UAV swarm network damage assessment model based on the entropy weight method-TOPSIS. Combined with actual battlefield environments, it establishes four primary evaluation metrics and thirteen secondary metrics: resilience, real-time performance, survivability, and communication assessment. The model evaluates the resilience of UAV swarm networks from four aspects: effectiveness, survivability, real-time performance, and communication recovery capability. Each primary metric includes multiple secondary metrics. For secondary metrics such as average clustering coefficient, betweenness, and toughness, this invention provides relevant evaluation methods, enabling the final score to achieve a comprehensive and high-precision evaluation of the network's resilience.

[0046] 3. Existing drone swarm network damage assessment methods only consider the node's own attributes when evaluating a single node, ignoring the influence of the attributes of its neighboring nodes on the node's importance. This invention draws on the concepts of exploration and policy in reinforcement learning to further consider the assessment model. In the previous model, the current node's score depended solely on its own attributes, such as betweenness and average shortest path ratio. However, in real-world scenarios, a node's importance often depends on the importance of its neighboring nodes, reflecting its pivotal role. Based on this idea, this invention proposes a comprehensive node importance scoring method based on the weighted correlation of neighboring node hop counts. This method correlates and accumulates a node's score with the scores of its surrounding neighboring nodes. By combining the accumulated score with the comprehensive score of the currently damaged node, the final score of the damaged node is obtained. The resulting score more accurately evaluates the network's resilience.

[0047] 4. Existing methods lack simulated battlefield environments and realistic battlefield missions, focusing solely on the UAV network topology. They lack realistic mission environments and metrics, and model metrics under battlefield mission-related conditions. Most studies assume little or no battlefield interference. The work of Li Yanjun et al., mentioned above, focuses on information and communication evaluation, lacking modeling and evaluation of real battlefield environments and missions. This invention establishes four primary evaluation metrics and thirteen secondary metrics to comprehensively evaluate the damage resistance of UAV swarm networks from four aspects: effectiveness, survivability, real-time performance, and communication recovery capability. All metrics at each level are constructed in conjunction with actual battlefield environments. Attached Figure Description

[0048] Figure 1 This invention provides a diagram of damage assessment indicators for unmanned aerial vehicle (UAV) swarm networks.

[0049] Figure 2 A simplified flowchart of the damage assessment method for UAV swarm networks based on the entropy weight method-TOPSIS provided by this invention;

[0050] Figure 3 The flowchart of the comprehensive evaluation of multiple neighbor nodes in the UAV swarm network damage assessment method based on entropy weight method-TOPSIS provided by the present invention. Detailed Implementation

[0051] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] This invention first studies the movement model, communication protocol, and routing protocol of UAV swarms. It decides to use an OLSR routing protocol and a biomimetic movement model in a WiFi environment for simulated damage simulation, and adds multiple network damage scenarios such as SIR fault propagation mechanism and cascading damage modeling, which are implemented in ns3. Secondly, this invention establishes a UAV swarm network damage assessment model based on the entropy weight method-TOPSIS. Combining actual battlefield environments, it establishes four primary evaluation indicators and thirteen secondary indicators to evaluate the damage resistance and resilience of the UAV swarm network from four aspects: effectiveness, survivability, real-time performance, and communication recovery capability.

[0053] Worm attacks are a common cybersecurity threat. Worms can self-replicate and spread copies across the network to other vulnerable hosts, thus expanding the scale of the infection. Worm attacks are particularly destructive in mobile ad hoc networks (such as drone swarms) because these networks are typically less secure and more vulnerable to worm attacks.

[0054] Cascading failure attacks refer to a situation where the failure of one or a few nodes triggers a chain reaction of failures in other nodes, potentially leading to the collapse of the entire network. In this attack pattern, each node carries a certain amount of load to perform tasks such as communication, forwarding, and migration. When a node's load exceeds its capacity due to a network attack, the node fails and distributes its load to its neighbors. If the neighboring nodes cannot handle this additional load, they may also fail, thus creating a cascading effect.

[0055] Drone swarms can perform both reconnaissance and attack missions. In an offensive scenario integrating reconnaissance and strike capabilities, drone swarms possess sustained and powerful strike capabilities, full coverage of the area, and the continuity of high-risk missions. This allows swarms composed of multiple drones of different or the same model to work together to complete missions, leveraging the advantages of drones while ensuring that issues such as the crash or poor performance of a single drone do not affect the successful completion of the entire mission.

[0056] To address the above issues, this invention provides a method for assessing damage to UAV swarm networks based on the entropy weight method-TOPSIS, specifically including the following steps:

[0057] S1: For the UAV swarm network, for the damaged nodes, obtain the values ​​of the secondary indicators under the four primary indicators of resilience, real-time performance, survivability and communication recovery capability. For each primary indicator, construct the indicator matrix corresponding to the primary indicator with its corresponding secondary indicator as the row and each damaged node as the column, and denoted as the primary indicator matrix.

[0058] Specifically, this invention primarily studies the movement model, communication protocol, and routing protocol of UAV swarms. It decides to use an OLSR routing protocol and a biomimetic movement model in a WiFi environment for simulated damage, and adds modeling of multiple network damage scenarios such as SIR fault propagation mechanism and cascading damage, all implemented in ns3. Secondly, this invention establishes a UAV swarm network damage assessment model based on the entropy weight method-TOPSIS. Combining this with the actual battlefield environment, it establishes four primary evaluation indicators and thirteen secondary indicators to evaluate the damage resistance and resilience of the UAV swarm network from four aspects: effectiveness, survivability, real-time performance, and communication recovery capability. The indicator classification of this invention is as follows: Figure 1 As shown, the primary indicator of resilience includes four secondary indicators: connectivity coefficient, average shortest path ratio, average clustering coefficient ratio, and betweenness coefficient. The average clustering coefficient is a network analysis metric used to describe the degree of clustering between nodes in a network. In a network, the clustering coefficient of each node is defined as the ratio of the possible number of connections between its neighbors to the actual number of connections. The average clustering coefficient is the sum of the clustering coefficients of all nodes. The mathematical formula can be expressed as:

[0059] The number of damaged nodes in the cluster network is N, and the clustering coefficient of the i-th damaged node is C. i The number of neighboring nodes of the damaged node i is k. i The actual number of connections between neighbors is E. i Then the clustering coefficient C of node i i for:

[0060] Betweenness centrality is a metric in network analysis that reflects the centrality of a node within a network. The more frequently a node appears in the shortest paths between all pairs of nodes, the higher its betweenness centrality. The betweenness centrality is the proportion of all shortest paths in the network that pass through a given node, and is calculated using the following formula:

[0061] Betweenness number B i The subscript i represents the i-th damaged node, i.e., B. i The betweenness of the damaged node i is specifically: Where g jk g is the number of shortest paths between damaged nodes j and k; jk (i) represents the number of shortest paths from j to k that pass through node i.

[0062] The primary metric, survivability, comprises four secondary metrics: connectivity coefficient, end-to-end reliability, k-terminal reliability, and overall reliability. Among these, k-terminal reliability is a network reliability metric used to describe the connectivity between k selected nodes in a network. Under different network conditions, k-terminal reliability reflects the probability that these nodes can maintain communication. Because it involves probability calculations for different network states, the calculation of k-terminal reliability is relatively complex and needs to be determined based on the actual network topology and conditions.

[0063] The primary metric, real-time performance, comprises four secondary metrics: throughput, transmission latency, packet loss rate, and average data transmission latency. Average data transmission latency is typically used to measure the efficiency of information propagation within a network. It refers to the average time required for information to travel from the source node to the destination node and back. It includes both data packet transmission time and network latency, and is a crucial indicator for evaluating network performance.

[0064] The primary indicator, communication recovery capability, includes a secondary indicator: resilience. In this embodiment of the invention, resilience describes the network's ability to maintain its original functionality or recover quickly when facing attacks or failures. Higher resilience indicates a stronger resistance to failures and a more robust recovery capability. Network resilience is typically measured by simulating attack or failure scenarios and observing changes in network performance.

[0065] Resilience assessment methods based on complex networks are a traditional type of resilience assessment method. They employ a comparison before and after a single damage-recovery event, where y is the value after the damage occurs. min Before the damage occurred y D The absorption factor δ was obtained by comparison, and the recovery occurred after y R Before the damage occurred y D The recovery factor ρ is obtained through comparison. This method focuses on the topology of the UAV swarm network itself, rather than the network's information exchange capabilities.

[0066] The resilience assessment method based on information interaction overcomes the shortcomings of the former. This method compares the ideal state of the swarm's information interaction capability with the state before recovery and the stable state after recovery following the damage event, thus accurately and reasonably evaluating the information interaction capability of the UAV swarm. Its resilience s is calculated as follows:

[0067] Resilience, specifically:

[0068] Toughness S is

[0069] Among them, σ, δ, ρ, τ, Let [t0, t] represent the system's performance factor, absorption factor, recovery factor, recovery event factor, and volatility factor, respectively. final The communication performance of the UAV within a given time period is y(t) as the basis; the calculation methods for each factor are as follows:

[0070]

[0071]

[0072]

[0073]

[0074]

[0075] The system's performance factor σ is obtained from the damage-recovery process and the undamaged process, used to measure the degree of performance degradation; the system's absorption factor δ is obtained from the damage before and after damage, used to measure the destructive force of the damage; the system's recovery factor ρ is obtained from the recovery after and before damage, used to measure the recovery effect; the system's recovery event factor τ is obtained from the start recovery time and total time, used to measure the recovery response speed; and the system's fluctuation factor... The fluctuation factor is obtained by combining the original signal and the smoothed signal, and is used to measure network fluctuation.

[0076] y min The average number of messages received from the end of damage to the start of recovery; y D y represents the average number of messages received before the damage begins. R The number of messages received after recovery; t ss To restore the start time; SNR dB Signal-to-noise ratio (SNR) in network communication, measured in decibels (dB).

[0077]

[0078]

[0079]

[0080] t th At the moment of destruction, t min At the moment of destruction, t ss To resume from the start time.

[0081] With the damaged nodes as columns and the secondary indicators as rows, establish a primary indicator matrix X = (x ij ) n×m Where n is the total number of secondary indicators corresponding to the primary indicator, m is the number of simulated damage nodes, and n = Kp A K p Each corresponds to the number of secondary indicators under a primary indicator, p = 1, 2, 3, 4. For example, when p = 1, it represents the primary indicator of resilience, n = 4, X = (x ij ) 4×m The corresponding indicators are four secondary evaluation indicators under resilience: connectivity coefficient, average shortest path ratio, average clustering ratio, and betweenness coefficient. The second dimension of the matrix is ​​the damaged node ID. When p=4, it represents the primary indicator communication evaluation, where n=1. The first row of the corresponding primary indicator matrix is ​​the resilience value, and the columns of the matrix are the damaged node IDs.

[0082] S2: Standardize each primary indicator matrix and then weight the standardized primary indicator matrix using a weight vector related to the entropy value to obtain the entropy-weighted standardized matrix of the primary indicators; specifically, this includes the following steps:

[0083] S201: The element x in the i-th row and j-th column of each first-level indicator matrix. ij This represents the value of the i-th secondary indicator under the primary indicator at the j-th damage node.

[0084] For each x ij Based on the optimization direction of their corresponding secondary indicators, they are classified and unified into the same optimization direction to obtain positively processed indicator values. Among them, connectivity coefficient, average clustering coefficient ratio, end-to-end reliability, K-segment reliability, overall reliability, and throughput are all positive indicators, while average shortest path ratio, betweenness, transmission delay, packet loss rate, and average round-trip time are negative indicators. The negative indicators are then subjected to range positiveization processing.

[0085] In multi-indicator evaluation, the optimization directions of each indicator may differ. Indicators are categorized according to their optimization direction and then unified towards a common optimization direction. Indicators with smaller values ​​generally have higher levels or quality; these are called negative indicators. Common methods for converting them into positive indicators include range positiveization and inverse proportional positiveization. Through this conversion, the original minimum value becomes the maximum value, and the maximum value becomes the minimum value, thus achieving positiveization.

[0086] Forwarding of range: x′ ij =max(x i )+main(x i )-x ij

[0087] Inverse proportional forward transformation: x′ ij =max(x i ) / x ij

[0088] Where x ijThis represents the performance of the i-th secondary indicator when the j-th node is damaged, x i Let x′ be the row vector corresponding to the i-th secondary index of matrix X. ij This represents x after forwarding. ij .

[0089] S202: Standardize the index values ​​obtained from the positive processing to obtain normalized index values; the specific method is as follows:

[0090]

[0091] Where, x ij Let x′ represent the value of the i-th metric in the UAV network when node j is damaged. ij After positive transformation, the replacement x ij This step primarily aims to remove the influence of unit dimensions. After processing, all indicators fall within the [0,1] interval. The normalized values ​​of each indicator are represented by X′. ij express.

[0092] S203: For the normalized index values, calculate the weight of each secondary index in the same secondary index at different damage nodes, where the weight of the i-th secondary index in the j-th damage node is denoted as P. ij ;

[0093]

[0094] Using this weight value, the entropy value of the secondary index is calculated, obtaining the entropy value of each secondary index corresponding to each of the four primary indices in the UAV network damage model, where the entropy value of the i-th secondary index is E. i :

[0095]

[0096] in m represents the total number of damaged nodes.

[0097] S204: Calculate the weight of the secondary indicator using the entropy value of the secondary indicator.

[0098]

[0099] Among them, G i G is the difference index for secondary indicator i. i =1-E i G i The larger the value, the greater the effect of the secondary indicator, and therefore the greater its corresponding weight.

[0100] The weight combination of secondary indicators belonging to the same primary indicator is the weight vector of that primary indicator, which is the entropy-related weight vector. The standardized primary indicator matrix is ​​weighted using the entropy-related weight vector to obtain the entropy-weighted standardized matrix of each primary indicator.

[0101] In this embodiment of the invention, the entropy-related weight vector for the primary metric of resilience is W = [W1, W2, W3, W4]. Taking resilience as an example, W1 represents the weight of connectivity coefficient in resilience, W2 represents the weight of average shortest path ratio in resilience, W3 represents the weight of average clustering ratio in resilience, and W4 represents the weight of betweenness coefficient in resilience. Each primary metric corresponds to an entropy-related weight vector.

[0102] For communication recovery capability, only the weight of resilience needs to be calculated, which is 1.

[0103] S3: Calculate the value score of each primary indicator based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicators; calculate the Euclidean distance D between each primary indicator and the positive and negative ideal solutions based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicators. - and D + D - and D + Let Euclidean distances be the best and worst values ​​of a secondary index for each node damaged, respectively, and the Euclidean distances between the best and worst values ​​of that secondary index when all nodes in the set of nodes are damaged sequentially.

[0104] The value score of a node in the current primary indicator is D. - / (D - +D + ).

[0105] This yields a score for the node based on primary metrics such as real-time performance, survivability, resilience, and communication recovery capability.

[0106] S4: For the value scores of each primary indicator, a comprehensive score for the damaged node is obtained by weighting the scores using a weight vector related to the entropy value. The obtained value scores of the primary indicators are then subjected to an entropy-weighted TOPSIS evaluation. The specific operation process is exactly the same as that for the secondary indicators, except that the object is the obtained primary indicator scores. After the same evaluation, the final evaluation score is obtained, which is the final evaluation score of this method for the network. The calculation process is as follows: Figure 2 As shown.

[0107] S5: Pre-set a hop count threshold, search for one-hop neighbor nodes for the current damaged node, and determine whether the hop count of the current neighbor node is less than the hop count threshold. If so, calculate and accumulate the sum of the comprehensive scores of the neighbor nodes with the current hop count. Multiply the accumulated score by a certain coefficient and then return to the hop count threshold for judgment. When the preset hop count threshold is reached, add the accumulated score to the comprehensive score of the current damaged node to obtain the final score of the damaged node.

[0108] This invention draws upon the concepts of exploration and strategy in reinforcement learning, further considering the evaluation model proposed earlier. In the previous model, the score of the current node depended solely on its own attributes, such as betweenness and average shortest path ratio. However, in real-world scenarios, the importance of a node often depends on the importance of its neighboring nodes, reflecting its pivotal role. Based on this idea, this invention proposes a comprehensive node importance scoring method based on the weighted correlation of neighboring node hop counts, thus linking a node's score with the scores of its surrounding neighboring nodes.

[0109] This invention, after obtaining node scores using the entropy weight method-TOPSIS, searches for, scores, and calculates neighboring nodes based on a pre-set hop count threshold. First, it determines whether the current neighbor hop count has reached the threshold, i.e., whether it exceeds the number of neighbors to be counted. Typically, it's appropriate to calculate two-hop neighbors in experiments; that is, the final score of a node should include a weighted sum of one-hop and two-hop neighbor scores. When the hop count hasn't reached the threshold, it calculates the sum of neighbor scores for the corresponding hop count in the next round, multiplies it by a certain coefficient, and then returns to the threshold judgment. When the threshold is reached, the previously accumulated score is added to the node's original TOPSIS score to obtain the node's final score. The purpose of this is to consider neighbor node attributes as an attribute of the node itself, fully considering the node's surrounding environment and making full use of its surrounding information. The process is as follows: Figure 3 As shown.

[0110] S6: The final score of all damaged nodes is used as the network damage assessment result for the drone swarm. A higher final score indicates that the node is more important. This invention combines the actual battlefield environment and clearly quantifies the assessment indicators under specific mission scenarios, which is of reference value for resisting attacks from today's complex drone swarms.

[0111] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing damage to UAV swarm networks based on the entropy weight method-TOPSIS, characterized in that, For unmanned aerial vehicle (UAV) swarm networks, for damaged nodes, the values ​​of secondary indicators under four primary indicators are obtained: resilience, real-time performance, survivability, and communication recovery capability. For each primary indicator, an indicator matrix is ​​constructed with its corresponding secondary indicators as rows and each damaged node as a column, denoted as the primary indicator matrix. Each primary indicator matrix is ​​standardized, and then weighted using a weight vector related to the entropy value to obtain the entropy-weighted standardized matrix of the primary indicators. This step is specifically executed as follows: S201: The element in the i-th row and j-th column of each first-level indicator matrix. This represents the value of the i-th secondary indicator under the primary indicator at the j-th damage node, for each... Based on the optimization direction of their corresponding secondary indicators, they are classified and unified into the same optimization direction to obtain positively processed indicator values. Among them, connectivity coefficient, average clustering coefficient ratio, end-to-end reliability, K-segment reliability, overall reliability, and throughput are all positive indicators, while average shortest path ratio, betweenness, transmission delay, packet loss rate, and average information transmission delay are negative indicators. The negative indicators are then subjected to range positive processing. S202: Standardize the index values ​​processed by positive transformation to obtain normalized index values; S203: For the normalized index values, calculate the weight of each secondary index in the same secondary index at different damage nodes, and use the weight value to calculate the entropy value of the secondary index, so as to obtain the entropy value of each secondary index corresponding to each primary index in the four primary indexes of the UAV network damage model. S204: Calculate the weight of the secondary indicator using the entropy value of the secondary indicator. The weight combination of the secondary indicators belonging to the same primary indicator is the weight vector of the primary indicator, which is the entropy-related weight vector. Use the entropy-related weight vector to perform weighting on the standardized primary indicator matrix to obtain the entropy-weighted standardized matrix of each primary indicator. Based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicator, calculate the value score of each primary indicator. For the value scores of each primary indicator, a comprehensive score for the damaged node is obtained by weighting the scores using a weight vector related to the entropy value. A hop count threshold is preset. The current damaged node is searched for one-hop neighbor nodes. It is determined whether the hop count of the current neighbor node is less than the hop count threshold. If so, the sum of the comprehensive scores of the neighbor nodes with the current hop count is calculated and accumulated. The accumulated score is multiplied by a certain coefficient and then returned to the hop count threshold for judgment. When the preset hop count threshold is reached, the accumulated score is added to the comprehensive score of the current damaged node to obtain the final score of the damaged node. The final score of all damaged nodes is used as the damage assessment result for the drone swarm network.

2. The method for assessing damage to UAV swarm networks based on entropy weight method-TOPSIS as described in claim 1, characterized in that, For the damaged nodes, the values ​​of secondary indicators under four primary indicators are obtained: resilience, real-time performance, survivability, and communication assessment. Specifically, the primary indicator of resilience includes the following four secondary indicators: connectivity coefficient, average shortest path ratio, average clustering ratio, and betweenness coefficient. The primary indicator of survivability includes the following four secondary indicators: connectivity coefficient, end-to-end reliability, K-terminal reliability, and global reliability. The primary indicator of real-time performance includes the following four secondary indicators: throughput, transmission latency, packet loss rate, and average information transmission latency. The primary indicator, communication recovery capability, includes the following secondary indicator: resilience.

3. The method for assessing damage to UAV swarm networks based on the entropy weight method-TOPSIS as described in claim 2, characterized in that, The average clustering coefficient, C, is specifically: ; The number of damaged nodes in the cluster network is N, and the clustering coefficient of the i-th damaged node is C. i The number of neighboring nodes of the damaged node i is k. i The actual number of connections between neighbors is E. i Then the clustering coefficient C of node i i for: .

4. The method for assessing damage to UAV swarm networks based on entropy weight method-TOPSIS as described in claim 2, characterized in that, The betweenness is B i The subscript i represents the i-th damaged node, i.e. B i The betweenness of the damaged node i is specifically: ;in This represents the number of shortest paths between damaged nodes j and k. This represents the number of shortest paths from node i to node k.

5. The method for assessing damage to UAV swarm networks based on entropy weight method-TOPSIS as described in claim 2, characterized in that, The toughness specifically refers to: Toughness S is in These represent the system's performance factor, absorption factor, recovery factor, recovery event factor, and volatility factor, respectively. The communication performance of the UAV within a given time period is y(t) as the basis; the calculation methods for each factor are as follows: The average number of messages received from the end of damage to the start of recovery; The average number of messages received before the damage began; This represents the number of messages received after recovery. To restore the start time; SNR dB Signal-to-noise ratio (SNR) in network communication, measured in decibels (dB). At the moment the damage begins, At the moment of destruction, To resume from the start time.

6. The method for assessing damage to UAV swarm networks based on entropy weight method-TOPSIS as described in claim 1, characterized in that, The calculation of the value score for each primary indicator based on the performance of each secondary indicator in the entropy-weighted standardized matrix of the primary indicator is as follows: According to the index performance of each secondary index in the entropy-weighted standardized matrix of the primary index, the Euclidean distance D of each primary index and the positive and negative ideal solutions is calculated - and D + , the value score of the node at the current primary index is D - / ( D - +D + ).

Images