A cluster self-organizing defense method of a protected target participating in cooperation

Abstract

The application relates to a kind of self-organizing defense methods of protected target participation cooperation, consider the two-way cooperation of defense cluster and protected target, wherein the self-organizing defense of protected target is realized based on improved social force model by defense cluster, contains three kinds of forces, such as intra-group interaction term, inter-group confrontation term and target defense term.Simultaneously, the situation awareness is realized based on improved dynamic Bayesian network by protected target, then corresponding strategy is selected to balance the demand of avoiding attack and reaching terminal point.The threat level is divided into three levels of low, medium and high, and the strategy corresponding to the three threat levels also reflects the degree of protected target participating in cooperative defense.Through the cooperation of self-organizing defense of defense cluster and cooperative defense of protected target, efficient defense can be finally realized.

Description

Technical Field

[0001] This invention belongs to the field of swarm intelligence technology and relates to a swarm self-organizing defense method in which protected targets participate in collaborative defense, and more particularly to a self-organizing control method that utilizes the situational awareness of protected targets to participate in swarm collaborative defense. Background Technology

[0002] Swarming is a spontaneous, emergent behavior at the group level, arising from the behavioral rules of individuals within a group. Examples include schools of fish swirling, flocks of birds hovering, and swarms of bees "flying." While the individuals forming these groups have limited capabilities, they exhibit complex, efficient, intelligent, and orderly self-organized emergent behaviors based on simple individual behavioral rules. Through observation, statistics, and analysis of biological behavior, researchers have proposed various classic theoretical models, such as the social force model, the Vicsek model, and the Cousin model. Swarming theory is also increasingly being applied to real-world scenarios, such as in drone swarms. Individual drones have limited capabilities and cannot perform complex and challenging tasks, while drone swarms can effectively overcome this limitation. Therefore, utilizing swarming technology to solve complex and demanding tasks has become a new option.

[0003] The widespread application of swarm technology in unmanned systems has also brought an unavoidable problem: the harassment of high-value targets by unmanned swarms. Since the attack strategies and scale of incoming swarms are completely unknown, it is impossible to pre-design specific countermeasures. Therefore, how to conduct self-organized defense against incoming swarms has become a key breakthrough in addressing this problem. Many studies have focused on this aspect, constructing various defense and adversarial models. While these studies have discussed strategies for defending swarms to counter incoming attackers, their strategy settings are more biased towards the pursuit and escape problem, without fully considering the collaborative strategies between high-value targets, i.e., protected targets, and the defending swarm.

[0004] Therefore, this invention proposes a self-organizing defense method involving the participation of the protected target in collaborative defense. It considers the bidirectional collaboration between the defense cluster and the protected target, whereby the defense cluster achieves self-organizing defense against the protected target based on an improved social force model. Simultaneously, the protected target achieves situational awareness based on an improved dynamic Bayesian network, and collaborates with the defense cluster to evade incoming attacks. Summary of the Invention

[0005] Technical problems to be solved

[0006] To avoid the shortcomings of existing technologies, this invention proposes a cluster self-organizing defense method involving the participation of protected targets in collaboration, ultimately achieving an effective two-way collaborative defense effect.

[0007] Technical solution

[0008] A cluster self-organizing defense method involving the participation of protected targets in collaboration is characterized by constructing a two-way collaborative defense model, namely, the defense cluster self-organizing defense method and the protected target collaborative defense method are used for defense.

[0009] The self-organizing defense method of the defense cluster: The defense cluster consists of N... D One defender D i Composed of, its position and velocity are x i and v i (i = 1, 2, ..., N) D The kinematic equations are:

[0010]

[0011] Where u i Indicates the defensive individual D i The resultant force consists of three forces:

[0012]

[0013] For group interaction items,

[0014] Group gatherings are completed through interactive elements within the group.

[0015] As an inter-swarm adversarial element, it defends against interactions between individuals within a cluster to achieve cluster aggregation and collision avoidance, and uses inter-swarm adversarial elements to block intrusions into the cluster.

[0016] For target protection items, defensive protection is provided to the protected target.

[0017] The protective target collaborative defense method performs the following defense steps:

[0018] Step 1: The protected target detects whether there is an attacker. If not, it is directly assessed as a low threat and proceeds directly to the endpoint. If an attacker exists, proceed to Step 2 and enter the threat level assessment process.

[0019] Step 2: Compare the number of defenders with the attacking force's numerical advantage n sup The nearest intruder is d attack And the recent intruder's heading angle deviation β attack Three-input target situational awareness dynamic Bayesian network structure;

[0020] Step 3: Input the conditional probability table and prior probability into the target situational awareness dynamic Bayesian network structure to obtain the posterior probability;

[0021] Step 4: The posterior probability vector output by the Bayesian network corresponds to the probability of occurrence of the three threat levels. The threat level with the highest probability is selected as the threat level at the current moment, and the posterior probability calculated at this moment is saved as the prior probability for the next moment.

[0022] Step 5: Select the appropriate strategy for coordinated defense:

[0023] When the threat level is low, the defense cluster can effectively protect the target, allowing it to maintain its original direction of travel and move towards the desired destination. The mathematical expression for this is shown below:

[0024]

[0025] Where v p and x p Let x represent the velocity and position of the protected target, respectively. goal Indicated as the expected endpoint of the protected target;

[0026] When the threat level is medium, the target actively moves towards the center of the defense cluster x. c To obtain protection, its mathematical expression is as follows:

[0027]

[0028] When the threat level is high, the target is extremely dangerous and needs to escape the attacker immediately. Its mathematical expression is as follows:

[0029]

[0030] Where T p d represents the set of attackers perceived by the protected object. jp This represents the distance from the attacking entity to the protected target.

[0031] The intra-group interaction is a local interaction, consisting of two parts: single-group aggregation and subgroup splitting, meaning it only interacts with neighbors within the perception range. The neighbor set is defined as follows:

[0032] N i ={k|d ik <R sen k∈{1,...,N D},k≠i}

[0033] Based on the location information of neighboring individuals within the sensing range, the force of interaction terms within the group is calculated:

[0034]

[0035] Where, d ik R represents the distance between a defending individual and its neighboring individuals.sen >0 indicates the perception radius of an individual within the defense cluster, N i Let ls represent the collision avoidance distance between individuals, μ represent the center distance of the repulsive force, and σ represent the distance between the centers of the repulsive force. 2 The range of influence of this repulsive force is defined. It is represented as the unit position vector from the neighboring individual to the defending individual.

[0036] The μ>ls.

[0037] The inter-group adversarial term refers to the pursuit and expulsion behavior of a defensive individual against an attacker within its perception range, and its mathematical expression is as follows: Among them, T i This represents the set of all attacking entities within the perception range. Where d... ik Indicates the distance between the defending entity and the attacking entity. It is represented as a unit position vector from the attacking individual to the defending individual.

[0038] The target defense term refers to the action of defensive individuals around the protected target, which manifests at the cluster level as the spontaneous formation of defensive formations. Mathematically, it is represented as follows: Where d ip Indicates the distance from the defending individual to the protected target. d represents the unit position vector from the defending individual to the protected target. warn This represents the defense radius of the defending individual.

[0039] The conditional probability table:

[0040]

[0041] When there is no attacker in the perception domain, the situation assessment result is directly low threat, and the prior probability is reset to the initial probability; when there is an attacker in the perception domain, the posterior probability of the previous moment is used as the prior probability of the current moment.

[0042] Beneficial effects

[0043] This invention proposes a cluster self-organizing defense method involving the participation of protected targets. It considers the bidirectional collaboration between the defense cluster and the protected target. The defense cluster achieves self-organized defense against the protected target based on an improved social force model, incorporating three forces: intra-cluster interaction, inter-cluster adversarial, and target defense. Simultaneously, the protected target achieves situational awareness based on an improved dynamic Bayesian network, and then selects appropriate strategies to balance the needs of attack avoidance and reaching the endpoint. Threat levels are divided into low, medium, and high, and the strategies corresponding to these three threat levels reflect the degree of participation of the protected target in collaborative defense. Through the combination of self-organized defense by the defense cluster and collaborative defense by the protected target, highly efficient defense can ultimately be achieved.

[0044] Beneficial effects:

[0045] 1. Due to the adoption of a self-organizing defense cluster method, the defense cluster can self-organize into a defense formation to protect the protected target and repel intruders, such as... Figure 4 As shown.

[0046] 2. Because the protection of the target participates in the coordination of the defense cluster, the defense capability is enhanced, such as... Figure 5 As shown, the protected target can adjust its behavioral strategies in a timely manner based on the threat assessment results and the current situation.

[0047] 3. By adding the step of judging whether there are neighbors within the perception range, the bias caused by the posterior probability remaining at the previous threat moment in the threat assessment results of dynamic Bayesian networks is corrected, thus improving the accuracy of the threat assessment results. Attached Figure Description

[0048] Figure 1 This is a flowchart of the collaborative defense algorithm for protected targets in the method of this invention. The defense cluster achieves intra-cluster aggregation through intra-cluster interaction terms, defends the protected target through target protection terms, and intercepts the intrusion cluster through inter-cluster adversarial terms. The protected target senses the presence of attackers; if none are present, it is directly assessed as a low threat and proceeds directly to the endpoint; if attackers are present, it enters the threat level assessment process. First, the number of defenders and the attacker's numerical advantage, the distance to the nearest intruder, and the heading angle deviation of the nearest intruder are input into the target situational awareness dynamic Bayesian network structure. A conditional probability table is constructed by combining expert knowledge and input along with prior probabilities into the network structure to obtain posterior probabilities, which are then used as the prior probabilities for the next time step. When the threat to the protected target is eliminated and no attackers are present within the sensing range, the prior probabilities of the network structure are initially reset to the initial probabilities. The obtained posterior probabilities correspond to three threat levels: low, medium, and high. When there is a high threat, a high-threat strategy is selected, in which the defense cluster intercepts the intruder while fleeing from the direction of the intruder's attack. When there is a medium threat, a medium-threat strategy is selected, in which the target moves closer to the defense cluster to seek its protection. When there is a low threat, a low-threat strategy is selected, in which the protected target moves directly to the endpoint without considering coordination.

[0049] Figure 2 This is a diagram of the dynamic Bayesian network structure for situational awareness of the protected target in the method of this invention.

[0050] Figure 3 This is a simulation interface diagram of the initial distribution of defense and countermeasures in the method of the present invention, which includes a defense cluster consisting of a group of 50 individual defenders, an attack (intrusion) cluster consisting of a group of 30 individual attackers, and a moving protected target.

[0051] Figure 4 This is a simulation diagram of the defender confronting the incoming target in the method of the present invention.

[0052] Figure 5 This is a simulation diagram of the collaborative defense of the protected target in the method of the present invention.

[0053] Figure 6 This is a simulation diagram showing the successful defense of the defensive cluster in the method of this invention. Detailed Implementation

[0054] The present invention will now be further described in conjunction with the embodiments and accompanying drawings:

[0055] This invention proposes a cluster self-organizing defense method capable of enabling protected targets to participate in collaborative defense, ultimately achieving an effective two-way collaborative defense effect. This cluster defense system has the following characteristics:

[0056] 1. Individuals within the defense cluster are isomorphic and follow the same behavioral rules. That is, the defense individuals are completely identical in all aspects, with no difference in size, shape, or individual capabilities.

[0057] 2. Individuals within a defensive cluster, i.e., defenders, possess a certain perception range. They can obtain information about the location and speed of neighboring individuals within this range, as well as the location of attackers. However, they have global perception of the protected target and can obtain its location information. Their task is to ensure the protected target reaches its desired endpoint.

[0058] 3. The protected target possesses a certain level of perception capability, enabling it to acquire information on the position and speed of defenders and attackers within its perception range. Its speed remains constant. Its mission is to reach the desired destination while avoiding capture by the attacker in the process.

[0059] 4. In defensive confrontation, capture or elimination is defined as the capture being completed and both the attacker and the target being eliminated simultaneously when the enemy enters the capture radius. A successful defense is defined as the capture of all incoming attackers within a specified time, or the successful protection of the target reaching the desired destination.

[0060] By introducing the above features, a two-way collaborative defense model was constructed, which includes two parts: a self-organizing defense method for defense clusters and a collaborative defense method for protected targets.

[0061] The self-organizing defense method of the defense cluster is based on an improved social force model, considering three roles: intra-cluster interaction, inter-cluster antagonism, and target protection. The defense cluster consists of N... D One defender D i The composition, its position and velocity can be expressed as: x i and v i (i = 1, 2, ..., N) D Its kinematic equations are as follows:

[0062]

[0063] Where u i Indicates the defensive individual D i The resultant force. It consists of three forces, and its mathematical expression is as follows:

[0064]

[0065] Among them, the group interaction items are represented as Intergroup adversarial term is represented as The target protection item is represented as The specific rule settings are as follows:

[0066] (1) Group interaction items

[0067] Intra-swarm interaction refers to the interaction between individuals within a defensive cluster to achieve clustering and collision avoidance. Intra-swarm interaction is localized, meaning it only interacts with neighbors within its perception range. The neighbor set can be defined as:

[0068] N i ={k|d ik <R sen k∈{1,...,N D},k≠i} (3)

[0069] Where, d ik R represents the distance between a defending individual and its neighboring individuals. sen >0 indicates the perception radius of an individual within the defense cluster.

[0070] Based on the location information of neighboring individuals within the sensing range, the force of interaction terms within the group can be calculated, i.e.:

[0071]

[0072] The intra-group interaction term consists of two parts: simple group aggregation and subgroup splitting. Where d ik N represents the distance between a defending individual and its neighboring individuals. i This represents the set of neighboring individuals. `ls` represents the collision avoidance distance between individuals. That is, when the distance from a neighboring individual to the defender is less than this value, a repulsive force is exhibited; when it is greater, an attractive force is exhibited, ultimately resulting in a single-cluster morphology at the cluster level. The last term introduces a Gaussian function to represent the repulsive force influence within a certain range. Here, `μ` represents the center distance of this repulsive force; that is, when the distance from a neighboring individual to the defender equals this value, the repulsive force reaches its maximum. Generally, `μ > ls`, meaning that the repulsive force appears after the collision avoidance distance, between long-range attraction and short-range repulsion. Therefore, this repulsive force is also called the medium-range repulsive force, which can enable the single-cluster morphology to split into a multi-subgroup morphology. And σ... 2The range of influence of this repulsive force is defined, which is expressed as a control parameter for the distance between subgroups at the cluster level. It is represented as the unit position vector from the neighboring individual to the defending individual.

[0073] (2) Intergroup rivalry

[0074] Inter-group adversarial behavior refers to the pursuit and expulsion actions of defensive individuals against attackers within their perception range. Its mathematical expression is shown below:

[0075]

[0076] Among them, T i This represents the set of all attacking entities within the perception range. Where d... ik Indicates the distance between the defending entity and the attacking entity. This is represented as the unit position vector from the attacking individual to the defending individual. Inter-swarm antagonism intensifies as the distance decreases; that is, the closer the attacking individual is to the defending individual, the stronger the defending individual's tendency to pursue it.

[0077] (3) Target Protection

[0078] The target defense term is the action term of defensive individuals around a protected target, defending it. At the cluster level, it manifests as the spontaneous formation of defensive formations. Its mathematical representation is shown below:

[0079]

[0080] Where d ip Indicates the distance from the defending individual to the protected target. d represents the unit position vector from the defending individual to the protected target. warn This represents the defense radius of an individual defender, i.e., the radius of the defense formation at the cluster level. This synergy controls the defense formation structure formed by the defense cluster. The larger the defense radius, the faster the threat is detected, while the more sparse the distribution of the defense cluster, thus this is also a key factor affecting defense efficiency.

[0081] Another way to further improve defense efficiency is through collaborative defense of the defense cluster by the protected target. To balance the needs of attack avoidance and reaching the endpoint, the collaborative defense method for protected targets proposed in this invention uses an improved dynamic Bayesian network to perceive the situation and select appropriate strategies to achieve reasonable defense. Its algorithm flowchart is as follows: Figure 1 As shown. The algorithm consists of the following steps:

[0082] Step 1: Select the main factors for situational assessment of the protected target;

[0083] Factors influencing the protected target include two aspects: defensive advantages and intrusive disadvantages. Defensive advantages refer to the numerical advantage (n) of defensive individuals relative to intrusive individuals within the perception range. sup Disadvantages of invasiveness include: the distance d between the most recently intruding individual and the target. attack And the heading angle deviation β between the most recently intruding individual and the target attack n sup The larger the size, the greater the defensive advantage, and the safer the protected target; d attack and β attack The larger the value, the less threat attackers pose to the protected target, and the safer the protected target is.

[0084] Step 2: Construct a dynamic Bayesian network structure for situation assessment of protected targets;

[0085] Taking into account the three factors mentioned above, a dynamic Bayesian network for assessing the situation of protected targets was constructed, such as... Figure 2 As shown.

[0086] Step 3: Provide a conditional probability table for each evaluation indicator based on expert theoretical knowledge;

[0087] Based on the statistical analysis of previous research data and combined with expert knowledge, a conditional probability table for this dynamic Bayesian network structure is presented, as shown in Table 1.

[0088] Table 1 Conditional Probability Table of Influencing Factors

[0089]

[0090] Step 4: Determine if an attacker exists within the perceptual domain;

[0091] This step is also an improvement on dynamic Bayesian networks. Traditional dynamic Bayesian networks use the posterior probability of the previous time step as the prior probability of the current time step, just like... Figure 2 As shown. However, when there is no attacker within the perception domain, the posterior probability from the previous moment will be retained as the prior probability for the current moment and even the next time an attacker is perceived. This characteristic can lead to a bias in the protected object's assessment of the current situation. Therefore, this step is added: when there is no attacker within the perception domain, the situation assessment result is directly low threat, and the prior probability is reset to the initial probability; when there is an attacker within the perception domain, the posterior probability from the previous moment is used as the prior probability for the current moment.

[0092] Step 5: Obtain the values ​​of each influencing factor at the current moment;

[0093] If step 4 determines that an attacker exists, then obtain the values ​​of the three influencing factors at the current moment and add them to the structure as evidence for the dynamic Bayesian network.

[0094] Step 6: Obtain the current threat level;

[0095] The posterior probability vector output by the Bayesian network is then obtained, which corresponds to the probability of occurrence of the three threat levels. The threat level with the highest probability is selected as the threat level at the current moment, and the posterior probability calculated at this moment is saved as the prior probability at the next moment.

[0096] Step 7: Select the appropriate strategy for coordinated defense.

[0097] The appropriate coordination strategy is selected based on different threat levels. Threat levels are divided into low, medium, and high. When the threat level is low, it indicates that the defensive forces are superior, and the defense cluster can effectively protect the target. In this case, the target maintains its original direction of travel towards the desired destination. Its mathematical expression is shown below:

[0098]

[0099] Where v p and x p Let x represent the velocity and position of the protected target, respectively. goal This represents the expected endpoint of the protected target.

[0100] When the threat level is medium, the target may be under attack, and the target will actively move towards the defense cluster center x. c To obtain protection, its mathematical expression is as follows:

[0101]

[0102] When the threat level is high, the target is extremely dangerous and needs to escape the attacker immediately. Its mathematical expression is as follows:

[0103]

[0104] Where T p d represents the set of attackers perceived by the protected object. jp This represents the distance from the attacking entity to the protected target.

[0105] Simulation results

[0106] The above algorithm was simulated. The main objects of the self-organizing defense method involving protected targets proposed in this invention include the protected targets and the defense cluster. In addition, a cluster invasion of a certain attack mode was introduced into the simulation, and the simulation effect is as follows. Figures 3-6 As shown. Among them. Figure 3 The initial distribution map for the defense confrontation includes a defense cluster consisting of a group of 50 defensive individuals, an attack (intrusion) cluster consisting of a group of 30 attacking individuals, and a moving protected target. Figure 4 This indicates that the defense cluster has formed a defense formation based on individual rules to deal with the incoming attack cluster. Figure 5 It reflects the response of the protected target to the current threat situation at a certain moment in the defensive confrontation process, while the defense cluster is protecting the protected target. Figure 6 The movement of the defense cluster after it has eliminated all attacking individuals is presented.

[0107] The simulation results demonstrate the effectiveness of the defense method and show that the protected object can accurately perceive the situation and make reasonable strategic choices.

Claims

1. A cluster self-organizing defense method involving the participation of protected targets in collaborative operations, characterized in that... Construct a two-way collaborative defense model, that is, to carry out defense using a self-organizing defense method for the defense cluster and a collaborative defense method for the protected target; The self-organizing defense method of the defense cluster: the defense cluster consists of... One defender Composition, its position and velocity are and The kinematic equations are: in Indicates a defensive individual The resultant force consists of three forces: For group interaction items, Group gatherings are completed through interactive elements within the group. As an inter-swarm adversarial element, it defends against interactions between individuals within a cluster to achieve cluster aggregation and collision avoidance, and uses inter-swarm adversarial elements to block intrusions into the cluster. For target protection items, defensive protection is provided to the protected target. The protective target collaborative defense method performs the following defense steps: Step 1: The protected target detects whether there is an attacker. If not, it is directly assessed as a low threat and proceeds directly to the endpoint. If an attacker exists, proceed to Step 2 and enter the threat level assessment process. Step 2: Utilize the numerical advantage of defenders against attackers. The nearest intruder was And the recent intruder's heading angle deviation Three-input target situational awareness dynamic Bayesian network structure; Step 3: Input the conditional probability table and prior probability into the target situational awareness dynamic Bayesian network structure to obtain the posterior probability; Step 4: The posterior probability vector output by the Bayesian network corresponds to the probability of occurrence of the three threat levels. The threat level with the highest probability is selected as the threat level at the current moment, and the posterior probability calculated at this moment is saved as the prior probability for the next moment. Step 5: Select the appropriate strategy for coordinated defense: When the threat level is low, the defense cluster can effectively protect the target, allowing it to maintain its original direction of travel and move towards the desired destination. The mathematical expression for this is shown below: in and These represent the speed and position of the protected target, respectively. Indicated as the expected endpoint of the protected target; When the threat level is medium, the target actively moves towards the center of the defense cluster. To obtain protection, its mathematical expression is as follows: ； When the threat level is high, the target is extremely dangerous and needs to escape the attacker immediately. Its mathematical expression is as follows: in This represents the set of attackers perceived by the protected object. This represents the distance from the attacking entity to the protected target.

2. The cluster self-organizing defense method for protected targets participating in collaboration according to claim 1, characterized in that: The intra-group interaction is a local interaction, consisting of two parts: single-group aggregation and subgroup splitting, meaning it only interacts with neighbors within the perception range. The neighbor set is defined as follows: Based on the location information of neighboring individuals within the sensing range, the force of interaction terms within the group is calculated: in, This indicates the distance between the defending individual and its neighboring individuals. This represents the perception radius of an individual within the defense cluster. Let ls represent the set of neighboring individuals, where ls represents the collision avoidance distance between individuals. When the distance from a neighboring individual to the defender is less than this value, a repulsive force is exhibited. This indicates the center distance of the repulsive force. The range of influence of this repulsive force is defined. It is represented as the unit position vector from the neighboring individual to the defending individual.

3. The cluster self-organizing defense method for protected targets participating in collaboration according to claim 2, characterized in that: The .

4. The cluster self-organizing defense method for protected targets participating in collaboration according to claim 1, characterized in that: The inter-group adversarial term refers to the pursuit and expulsion behavior of a defensive individual against an attacker within its perception range, and its mathematical expression is as follows: ,in, Represented as the set of all attacking entities within the perception range, where Indicates the distance between the defending entity and the attacking entity. It is represented as a unit position vector from the attacking individual to the defending individual.

5. The cluster self-organizing defense method for protected targets participating in collaborative efforts according to claim 1, characterized in that: The target defense term refers to the action of defensive individuals around the protected target, which manifests at the cluster level as the spontaneous formation of defensive formations. Mathematically, it is represented as follows: ,in Indicates the distance from the defending individual to the protected target. This represents the unit position vector from the defending individual to the protected target. This represents the defense radius of the defending individual.

6. The cluster self-organizing defense method for protected targets participating in collaboration according to claim 1, characterized in that: The conditional probability table: 。 7. The cluster self-organizing defense method for protected targets participating in collaboration according to claim 1, characterized in that: When there is no attacker in the perception domain, the situation assessment result is directly low threat, and the prior probability is reset to the initial probability; when there is an attacker in the perception domain, the posterior probability of the previous moment is used as the prior probability of the current moment.

Images