Goal-oriented decentralized unmanned swarm dynamic formation method, equipment, medium and product for fragmented information

Abstract

The method, equipment, medium and product of decentralized unmanned cluster dynamic formation for target-oriented fragmented information belong to the field of unmanned equipment control technology, and solve the problem that it is difficult to integrate multiple single methods for solving a certain type of problem to complete cluster tasks, and that unmanned cluster intelligent devices cannot effectively complete tasks when information is incomplete. The method of the present invention includes: defining a subgroup expression method that can characterize the relationship between devices in an unmanned cluster; defining a local reachable density and an outlier index, and realizing subgroup reconstruction through the separation of subgroups to achieve the effect of cluster dynamic formation, and completing target allocation at the same time; guided by the task goal, defining the concept of separation index based on the local reachable density, and realizing decentralized unmanned cluster dynamic formation through the aggregation and separation of unmanned subgroups. The present invention is suitable for dynamic formation when unmanned clusters work collaboratively to complete collaborative tasks.

Description

Technical Field

[0001] The present application relates to the field of unmanned equipment control technology, and in particular to decentralized unmanned cluster dynamic formation. Background Art

[0002] Decentralized unmanned swarms are becoming a hot topic in the development of small unmanned equipment. Unmanned swarm path planning, swarm formation, swarm communication and decentralization are important issues that need to be solved in the field of unmanned swarms. As far as unmanned equipment formation is concerned, scholars have proposed various formation algorithms, such as virtual structure method, leader-follower method and behavior-based method; for path planning problems, existing solutions include artificial potential field method, neural network method, optimization theory algorithm and collaborative formation dynamic obstacle avoidance control algorithm.

[0003] At present, scholars tend to focus on single research to solve path planning, formation methods or decentralization problems. However, when facing practical engineering problems in specific application scenarios, it is difficult to integrate multiple single methods to solve a certain type of problem to complete cluster tasks. In addition, with the increase in the size of the cluster and the number of targets, as well as complex environments such as the mission execution area and communication interference, unmanned cluster intelligent devices may face fragmented information such as communication interruption, incomplete information or even incomplete information during the execution of tasks, which puts higher requirements on unmanned cluster control. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that it is difficult to integrate multiple single methods for solving a certain type of problem to complete cluster tasks and that unmanned cluster intelligent devices cannot effectively complete tasks when information is incomplete. It provides a goal-oriented decentralized unmanned cluster dynamic formation method, equipment, medium and product for fragmented information.

[0005] The present invention is implemented by the following technical solutions. On the one hand, the present invention provides a goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information, the method comprising:

[0006] Step 1: Divide the unmanned cluster into subgroups consisting of several unmanned devices;

[0007] Step 2: Obtain unmanned device description information, unmanned device subgroup description information, and the attraction and repulsion between devices in the unmanned cluster;

[0008] Step 3: Obtain the local reachable density of each device in the unmanned cluster subgroup. The local reachable density represents the individual density within the θ neighborhood with the unmanned device O as the center. Point O represents a certain unmanned device and its location in the subgroup, and the θ neighborhood is the maximum self-perception distance of the unmanned device in the absence of communication.

[0009] Step 4: According to the local reachable density and the attraction and repulsion between individuals, the local outlier index is obtained. The calculation formula of the local outlier index is:

[0010]

[0011] Among them, P i (O)∈K k (O), K k (O) represents the set of individuals within the kth reachable distance from individual O, |K k (O)| represents the size of the set; f(O,j) is the attraction and repulsion of individual O in the subgroup by individual j; α and β are artificially set weight coefficients, which can be adjusted according to the size of the unmanned cluster and the specific problems it faces; lrded k (O) represents the kth local reachable density centered at O;

[0012] Step 5, determine the individual behavior according to the local outlier index in step 4; when the local outlier index is greater than the specified threshold, the individual leaves the original subgroup and is attracted to the subgroup where the neighboring individual of the non-same subgroup with the greatest attraction and repulsion is located; when there are no other subgroups around it, the individual separates from the original subgroup and becomes a new subgroup independently;

[0013] Update the triplet information representing the subgroup status corresponding to the unmanned device individual, and update the subgroup information in the individual attributes;

[0014] Step 6: By updating the status of the sub-groups and the status of the individual unmanned devices, the overall status of the unmanned cluster is updated to achieve target-oriented dynamic formation of each sub-group.

[0015] Furthermore, the unmanned equipment description information includes: theoretical target number, perceived target number, unmanned equipment number, coordinates, speed, direction and number of individuals in the subgroup;

[0016] The unmanned device subgroup description information includes: the unmanned device numbers contained in the subgroup, the virtual edges connecting the devices in the unmanned subgroup, and the degree-value function related to degree in the unmanned subgroup network.

[0017] Furthermore, the function of degree-degree correlation in the unmanned subgroup network is:

[0018]

[0019] Among them, d ij represents the distance between unmanned equipment individual i and the individual j closest to i, ρ σ (i) represents the local reachability density of individual i in the neighborhood of σ, which is used to describe the number of individuals in the subgroup, θ 1 and θ2 is the coefficient, θ 1 +θ 2 =1, σ is a constant parameter, which can be set according to the characteristics of the unmanned cluster and is used to describe the range that the unmanned equipment can perceive on its own.

[0020] Furthermore, the calculation formula of the attraction and repulsion between the devices in the unmanned cluster is:

[0021]

[0022]

[0023] Among them, K(i) and K(j) represent the degrees of unmanned equipment individuals i and j respectively. Assume that the initial relative position coordinates of the unmanned cluster are (0,0), and the relative position coordinates of unmanned equipment i at time q are (x i,q ,y i,q ),d ij is the Euclidean distance between device individuals i and j, k is a constant coefficient, and is adaptively adjusted according to the length of the unmanned device experimental scene.

[0024] Furthermore, the kth local reachable density lrded centered at O k The formula for (O) is:

[0025]

[0026] Among them, K k (O) represents the set of individuals within the kth reachable distance from individual O, |K k (O)| represents the size of the set; the formula for the kth reachable distance is:

[0027]

[0028] Among them, θ is a parameter, which is set to the maximum reachable distance. When P is unmanned, it indicates o and P j Unreachable; d k(O) Subgroup cluster center point P O To point P k(O) The distance between individuals in the subset and O is sorted, P k(O) is the kth closest point to O; d (O,Pj(O)) Represents unmanned equipment individuals O and P j The distance between j is the j-th closest point to point O.

[0029] Furthermore, θ is taken as the maximum distance that each unmanned device can perceive itself without communication.

[0030] Further, in step 5, the triplet information is S(C MN ), including:

[0031] Subgroup C MN The state is represented by S(C MN ), which can be described as a triple, storing: the number of individuals in the subgroup; the status of each unmanned device in the current subgroup; the status of the subgroup's corresponding target; where N is the number of unmanned devices in the subgroup, and M is the number of tasks that the subgroup needs to perform. Individuals with the same tasks are considered to be in the same subgroup, and the number of individuals in the subgroup is calculated in this way.

[0032] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the steps of a goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information as described above are executed.

[0033] In a third aspect, the present invention provides a computer-readable storage medium, in which a plurality of computer instructions are stored, and the plurality of computer instructions are used to enable a computer to execute a goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information as described above.

[0034] In a fourth aspect, the present invention provides a computer program product, which, when executed by a processor, implements a goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information as described above.

[0035] Beneficial effects of the present invention:

[0036] Complex unmanned swarm application scenarios place higher demands on their swarm formation technology. Faced with dynamic application scenarios, limited communications, incomplete information, or even only a small amount of fragmented information, it is necessary to achieve swarm coordination based on the limited information perceived by the unmanned equipment itself and complete multiple specified tasks.

[0037] The present invention proposes a goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information. Guided by the task goal, the concept of separation index is defined based on the local reachable density, and decentralized unmanned cluster dynamic formation is realized through the aggregation and separation of unmanned sub-groups. It solves the cluster formation problem in situations such as incomplete information and fragmented information caused by limited cluster communication, and realizes efficient decentralized unmanned cluster collaborative operation.

[0038] The goal-oriented decentralized unmanned swarm dynamic formation method for fragmented information can solve the dynamic formation of unmanned swarms in large-scale, dynamic, and incomplete information situations, allowing the swarms to collaboratively complete multiple specified tasks. This method can be used for the collaborative operation of large-scale unmanned swarm equipment such as drones and unmanned vehicles under fragmented information.

[0039] The present invention is applicable to the dynamic formation of unmanned swarms working in large-scale, communication-restricted, complex environments and incomplete information conditions to complete collaborative tasks. It can be used for the collaborative operation of large-scale unmanned swarm equipment such as drones and unmanned vehicles under fragmented information. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0041] Figure 1 A schematic diagram of the relationship between the attraction and repulsion forces between individuals of the present invention;

[0042] Figure 2 is the kth distance of the present invention (when k=5, d 5 (O)) Schematic diagram;

[0043] Figure 3 The schematic diagram of the reachable distance and range of the individuals of the present invention with each individual as the center is as follows: the largest circle represents the range of the kth distance of individual O when k=3, and the individuals within the range;

[0044] Figure 4 The present invention is 8 (O,P 8 ) is essentially a schematic diagram of the 7th reachable distance;

[0045] Figure 5 The point P of the present invention is relative to O 1 and O 2 A schematic diagram of the location of and the subgroup to which it belongs;

[0046] Figure 6 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0047] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0048] Specific implementation method 1: a target-oriented decentralized unmanned swarm dynamic formation method for fragmented information, the method comprising:

[0049] Step 1: Divide the unmanned cluster into subgroups consisting of several unmanned devices;

[0050] Step 2: Obtain unmanned device description information, unmanned device subgroup description information, and the attraction and repulsion between devices in the unmanned cluster;

[0051] Step 3: Obtain the local reachable density of each device in the unmanned cluster subgroup. The local reachable density represents the individual density within the θ neighborhood with point O as the center. Point O represents a specific unmanned device in the subgroup, and the θ neighborhood is the distance at which the device can perceive external information. The maximum self-perception distance of the unmanned device in the absence of communication is the maximum reachable distance.

[0052] It should be noted that, because it is a scenario facing fragmented information, each unmanned device does not fully understand the information of other individuals in the cluster. Therefore, the behavior of each unmanned device is based on its own information, so each calculates its own local reachable density. For each unmanned device with the same goal, it is considered to be in the subgroup of the same unmanned cluster.

[0053] Step 4: According to the local reachable density and the attraction and repulsion between individuals, the local outlier index is obtained. The calculation formula of the local outlier index is:

[0054]

[0055] Among them, P i (O)∈K k (O),|K k (O)| represents the size of the set; f(O,j) is the attraction and repulsion of individual O in the subgroup by individual j; α and β are artificially set weight coefficients, which can be adjusted according to the size of the unmanned cluster and the specific problems it faces; lrded k (O) represents the kth local reachable density centered at O.

[0056] Step 5: According to the local outlier index in step 4, when the outlier index is greater than the specified threshold, the individual leaves the original subgroup and is attracted to the subgroup where the neighboring individual of the non-same subgroup with the greatest attraction and repulsion is located; when there are no other subgroups around it, the individual separates from the original subgroup and becomes a new subgroup independently;

[0057] Update the triplet information representing the subgroup status corresponding to the unmanned device individual, and update the subgroup information in the individual attributes;

[0058] Step 6: By updating the status of the sub-groups and the status of the individual unmanned devices, the overall status of the unmanned cluster is updated to achieve target-oriented dynamic formation of each sub-group.

[0059] In this embodiment, unmanned equipment may include unmanned aerial vehicles, unmanned boats, unmanned vehicles and other intelligent devices with autonomous behavior capabilities. Fragmented information refers to the information that remains after information loss due to poor communication, technical limitations, malicious damage, etc. in the application scenario, and it is difficult to understand the status of the unmanned cluster as a whole, or to perform scheduling planning, etc. based on the retained information. The entire unmanned cluster is regarded as consisting of several unmanned cluster subgroups, and each unmanned device in the subgroup is called an individual. The unmanned cluster subgroups are composed of devices with the same goal, that is, the individuals in each subgroup have the same goal.

[0060] This implementation method is guided by the task objectives and defines the concept of separation index based on the local reachable density. It realizes the decentralized unmanned cluster dynamic formation through the aggregation and separation of unmanned sub-groups, solves the cluster formation problem in the case of incomplete information and fragmented information caused by limited cluster communication, and realizes efficient decentralized unmanned cluster collaborative operation.

[0061] In the case of incomplete and poorly-smoothed information in unmanned swarm mission scenarios, how to achieve adaptive formation of the swarm and complete multiple specified tasks when each unmanned device can only sense limited surrounding information has been a difficult research issue in recent years.

[0062] A goal-oriented decentralized unmanned swarm system adaptive formation method for fragmented information is proposed. A subgroup separation method based on outlier index is proposed to realize the reconstruction of each subgroup in the cluster and achieve the formation effect. The goal-oriented subgroup separation enables this method to simultaneously determine the target of each device in each subgroup.

[0063] This method can solve the problem of multi-target allocation in unmanned clusters when the individual information in the cluster is incomplete or fragmented. Specifically, it includes:

[0064] Defines a subgroup expression method that can characterize the relationship between devices in an unmanned cluster;

[0065] The local reachable density and outlier index are defined, and subgroup reconstruction is achieved through subgroup separation to achieve the effect of cluster dynamic formation and complete target allocation at the same time.

[0066] Specific implementation method 2: This implementation method is a further limitation of the target-oriented decentralized unmanned cluster dynamic formation method for fragmented information described in implementation method 1. In this implementation method, the description of the unmanned device and the description of the unmanned device subgroup are further limited, specifically including:

[0067] The unmanned equipment description information includes: theoretical target number, perceived target number, unmanned equipment number, coordinates, speed, direction and number of individuals in the subgroup;

[0068] The unmanned device subgroup description information includes: the unmanned device numbers contained in the subgroup, the virtual edges connecting the devices in the unmanned subgroup, and the degree-degree related function in the unmanned subgroup network.

[0069] In this implementation, the formation of starling flocks is used as a reference to formally describe each device (individual) in the unmanned cluster, so that the information of each individual in the cluster can be converted into mathematical language to describe the cluster, laying the foundation for the mathematical description of the cluster in the cluster formation algorithm.

[0070] In cluster formation, mathematical language is needed to describe the information of each individual in the cluster, as well as the information of the cluster as a whole, so as to realize the description of the cluster formation algorithm.

[0071] Specific implementation method three, this implementation method is a further limitation of the target-oriented decentralized unmanned cluster dynamic formation method for fragmented information described in implementation method two. In this implementation method, the degree-degree related function in the unmanned subgroup network is further limited, specifically including:

[0072] The function of degree-degree correlation in the unmanned subgroup network is:

[0073]

[0074] Among them, d ij represents the distance between unmanned equipment individual i and the individual j closest to i, ρ σ (i) represents the local reachability density of individual i in the neighborhood of σ, which is used to describe the number of individuals in the subgroup, θ 1 and θ 2 is the coefficient, θ 1 +θ 2 =1, σ is a constant parameter, which can be set according to the characteristics of the unmanned cluster and is used to describe the range that the unmanned equipment can perceive on its own.

[0075] The degree-degree correlation function in the unmanned subgroup network in this embodiment is used to characterize the connection / disconnection trend between the kth unmanned device and the adjacent individuals (nodes). The change in degree value is related to the distance between each node and the number of individuals in the subgroup. Nodes with small degrees tend to connect to other nodes, which is called positive incidence; nodes with large degrees tend to disconnect from other nodes, which is called negative incidence. In the subgroup, the degree of each node must be greater than the specified domain value (such as K(i)>0.5) to avoid the unmanned devices being too close.

[0076] Specific implementation method 4: This implementation method is a further limitation of the target-oriented decentralized unmanned cluster dynamic formation method for fragmented information described in implementation method 3. In this implementation method, the attraction and repulsion between the devices in the unmanned cluster are further limited, specifically including:

[0077] The calculation formula of the attraction and repulsion between the devices in the unmanned cluster is:

[0078]

[0079]

[0080] Among them, K(i) and K(j) represent the degrees of unmanned equipment individuals i and j respectively. Assume that the initial relative position coordinates of the unmanned cluster are (0,0), and the relative position coordinates of unmanned equipment i at time q are (x i,q ,y i,q ),d ij is the Euclidean distance between device individuals i and j, k is a constant coefficient, and is adaptively adjusted according to the length of the unmanned device experimental scene.

[0081] In this embodiment, the position adjustment pattern of individuals in a starling cluster is simulated, and the attraction and repulsion force f(i, j) between nodes is introduced to express the relationship between individual nodes. Based on this, the autonomous evolution of the cluster structure and spatial changes is analyzed, and a description of the cluster is formed through the description of multiple individuals.

[0082] Consider a starling flock as an unmanned cluster, where each starling corresponds to an unmanned device, and the behavior of each starling is determined by the behavior of its neighboring starlings and the surrounding conditions it observes. Birds at the edge of the starling flock are more likely to detect the surrounding environment and respond to the target (food or natural enemy) under the pull of the target: fly towards or fly away.

[0083] Here we describe the relationship between individuals and their neighboring individuals, and assume that there is no influence from the surrounding environment at present, so as to describe the unmanned cluster network.

[0084] like Figure 1 As shown, individual i first identifies individual j according to its n nearest neighbors (neighbors) 1 ,j 2 ,j 3 ,j 4 The attraction and repulsion information of each individual determines its position in the network, while considering its n distant neighbors (distant neighbors) individual k 1 ,k 2 ,k 3 ,k 4The attraction and repulsion of individual i and its four neighbors are affected by 2*n, 2*n is the number of neighbors, usually defined as n = 4. 1 ),f(i,j 2 ),f(i,j 3 ),f(i,j 4 ) reaches a dynamic equilibrium, that is, when the attraction and repulsion between individual i and its neighboring individuals are 0, the position of i is the optimal position of the individual in the network.

[0085] For the attraction and repulsion f(i,j) between individuals i and j, it is related to the distance between the individuals and is defined as:

[0086] 1) When the attraction and repulsion f(i,j) = 0, individuals i and j are at the appropriate distance d;

[0087] 2) When the attraction and repulsion f(i,j)>0, the distance between individuals i and j is too close, and the two individuals will swim back to back and distance themselves from each other. At this time, the attraction and repulsion manifests as repulsion;

[0088] 3) When the attraction and repulsion f(i,j)<0, the distance between individuals i and j is too far, and the two individuals will swim towards each other and get closer. At this time, the attraction and repulsion manifests as attraction.

[0089] Specific implementation method 5: This implementation method is a further limitation of the target-oriented decentralized unmanned cluster dynamic formation method for fragmented information described in implementation method 4. In this implementation method, the local reachable density is further limited, specifically including:

[0090] O represents a specific unmanned device location, and the kth local reachable density lrded centered at 0 k The formula for (O) is:

[0091]

[0092] Among them, K k (O) represents the set of individuals within the kth reachable distance from individual O, |K k (O)| represents the size of the set; the formula for the kth reachable distance is:

[0093]

[0094] Among them, θ is a parameter, which is set to the maximum reachable distance. When P is unmanned, it indicates o and P j Unreachable; d k(O) Subgroup cluster center point P O To point P k(O) The distance between individuals in the subset and O is sorted, Pk(O) is the kth closest point to O; d (O,Pj(O)) Represents unmanned equipment individuals O and P j The distance between j is the point (or set of points) that is closest to point O.

[0095] In this implementation, the local reachable density is used to characterize the density of individuals around point O. When the local reachable density increases, it means that the number of individuals in the current range increases, and the outlier index of the individuals in the range increases, otherwise the outlier index decreases. When the subgroups aggregate, the local reachable density increases, but cannot be greater than a specified threshold to avoid collision.

[0096] Specific implementation method 6: This implementation method is a further limitation of the target-oriented decentralized unmanned cluster dynamic formation method for fragmented information described in implementation method 5. In this implementation method, the value of θ is further limited, specifically including:

[0097] The value of θ is the maximum distance that each unmanned device can perceive itself without communication.

[0098] The larger the value of θ is, the stronger the perception capability of the unmanned device is, and the more information it can know about other devices in the cluster.

[0099] Specific implementation method seven, this implementation method is a further limitation of the target-oriented decentralized unmanned cluster dynamic formation method for fragmented information described in implementation method one. In this implementation method, the triple information in step 5 is further limited, specifically including:

[0100] In step 5, the triple information is S(C MN ), including:

[0101] Subgroup C MN The state is represented by S(C MN ), which can be described as a triple, storing: the number of individuals in the subgroup; the status of each unmanned device in the current subgroup; the status of the subgroup's corresponding target; where N is the number of unmanned devices in the subgroup, and M is the number of tasks that the subgroup needs to perform. Individuals with the same tasks are considered to be in the same subgroup, and the number of individuals in the subgroup is calculated in this way.

[0102] In this embodiment, the state S(C MN ) is related to the number of individuals in the subgroup and the target completed workload of the current subgroup. The completed workload of the current subgroup is represented by the working time and the number of individuals in the subgroup.

[0103] Specific implementation method eight, this implementation method is an implementation example of a target-oriented decentralized unmanned cluster dynamic formation method for fragmented information as described above, specifically including: Figure 6 shown.

[0104] The entire unmanned cluster is considered to be composed of several unmanned cluster sub-groups, and each unmanned device in the sub-group is called an individual.

[0105] Step 1: Drawing on the formation of starling flocks, a formal description of each device (individual) in the unmanned cluster is made, so that the information of each individual in the cluster can be converted into mathematical language to describe it, laying the foundation for the mathematical description of the cluster in the cluster formation algorithm.

[0106] In cluster formation, mathematical language is needed to describe the information of each individual in the cluster, as well as the information of the cluster as a whole, so as to realize the description of the cluster formation algorithm.

[0107] An unmanned cluster is composed of several unmanned devices. During the execution of the mission, the unmanned cluster is divided into subgroups consisting of one or several individuals. The number of unmanned devices in each subgroup changes dynamically according to the attraction and repulsion between the devices and the targets found. The subgroups are combined together to form an unmanned cluster. Based on this, the state of the entire unmanned cluster can be explained by a formal description of the subgroups, that is, the unmanned cluster contains several subgroups, each subgroup contains several unmanned devices, and the relationship between the unmanned devices is expressed by attraction and repulsion. The information of each unmanned device includes the current position, flight speed and direction. The dynamic change process of each subgroup, including the change in the number of devices and the information of each unmanned device, is the process of dynamic formation of the unmanned cluster.

[0108] To this end, the relationship between unmanned individuals (attraction and repulsion), unmanned sub-groups, and unmanned equipment individuals are described separately to describe the unmanned cluster.

[0109] (1) Repulsive force:

[0110] The position adjustment pattern of individuals in a starling cluster is simulated, and the attraction and repulsion force f(i, j) between nodes is introduced to express the relationship between individual nodes. Based on this, the autonomous evolution of cluster structure and spatial changes is analyzed, and a description of the cluster is formed through the description of multiple individuals.

[0111] Consider a starling flock as an unmanned cluster, where each starling corresponds to an unmanned device, and the behavior of each starling is determined by the behavior of its neighboring starlings and the surrounding conditions it observes. Birds at the edge of the starling flock are more likely to detect the surrounding environment and respond to the target (food or natural enemy) under the pull of the target: fly towards or fly away.

[0112] Here we describe the relationship between individuals and their neighboring individuals, and assume that there is no influence from the surrounding environment at present, so as to describe the unmanned cluster network.

[0113] For a starling flock in free space, a starling (individual) on the plane is set with several neighbor nodes, such as Figure 1 For the convenience of description, the number of neighbor nodes is taken as 8.

[0114] exist Figure 1 In the example, individual i first identifies individual j according to its n nearest neighbors (neighbors) 1 ,j 2 ,j 3 ,j 4 The attraction and repulsion information is used to determine its position in the network, while considering its n more distant (i.e., individuals from n+1 to n+n) neighbors (distant neighbors) individuals k 1 ,k 2 ,k 3 ,k 4 The attraction and repulsion of individual i and its four neighbors are affected by 2*n, 2*n is the number of neighbors, usually defined as n = 4. 1 ),f(i,j 2 ),f(i,j 3 ),f(i,j 4 ) reaches a dynamic equilibrium, that is, when the attraction and repulsion between individual i and its neighboring individuals are 0, the position of i is the optimal position of the individual in the network.

[0115] For the attraction and repulsion f(i,j) between individuals i and j, it is related to the distance between the individuals and is defined as:

[0116] 1) When the attraction and repulsion f(i,j) = 0, individuals i and j are at the appropriate distance d;

[0117] 2) When the attraction and repulsion f(i,j)>0, the distance between individuals i and j is too close, and the two individuals will swim back to back and distance themselves from each other. At this time, the attraction and repulsion manifests as repulsion;

[0118] 3) When the attraction and repulsion f(i,j)<0, the distance between individuals i and j is too far, and the two individuals will swim towards each other and get closer. At this time, the attraction and repulsion manifests as attraction.

[0119] (2) Description of unmanned equipment subgroups

[0120] Graph theory and complex network theory are used to formally describe the subgroups in the dynamic unmanned cluster.

[0121] A network model consisting of triples (node ​​set, edge set, degree value function set) is designed to express and describe the subgroup configuration of the dynamic unmanned cluster and its related implicit order and action rules, forming an expression that can describe the characteristics of the unmanned cluster. An unmanned equipment cluster consists of several unmanned equipment, and unmanned equipment with the same goal constitutes an unmanned equipment subgroup. The cluster is expressed as:

[0122] G=(V,E,K(i))

[0123] in:

[0124] 1) V = {1, 2, 3, ..., n} is a set of n nodes, n is the number of devices in the unmanned subgroup, and also represents the number of unmanned devices;

[0125] 2) E = {(i,j)|i,j∈V} is a set of m undirected edges, representing the virtual edges connecting the devices in the unmanned subgroup, and i and j represent the i-th and j-th devices in the unmanned subgroup, respectively.

[0126] 3) K(i) is the degree-correlated degree function in the unmanned subgroup network. It is used to characterize the connection / disconnection trend between the kth unmanned device and the adjacent individuals (nodes). The change in degree value is related to the distance between each node and the number of individuals in the subgroup. Nodes with small degrees tend to connect to other nodes, which is called positive incidence; nodes with large degrees tend to disconnect from other nodes, which is called negative incidence. In the subgroup, the degree of each node must be greater than the specified domain value (such as K(i)>0.5) to avoid the unmanned devices being too close. The formula for K(i) is described as follows:

[0127]

[0128] where d ij represents the distance between unmanned equipment individual i and the individual j closest to i, ρ σ (i) represents the local reachability density of individual i in the neighborhood of σ, which is used to describe the number of individuals in the subgroup, θ 1 and θ 2 is the coefficient, θ 1 +θ 2 =1, σ is a constant parameter, which can be set artificially according to the characteristics of the unmanned cluster and is used to describe the range that the unmanned equipment can perceive on its own.

[0129] (3) Unmanned equipment

[0130] The unmanned device is represented as a seven-tuple, which includes: theoretical target number, perceived target number, unmanned device number, coordinates, speed, direction, and the number of individuals in the subgroup.

[0131] Among them, the theoretical target number is the target when the unmanned device starts, which is used to determine the subgroup where the device is initially located; the perceived target number is the target perceived by the individual during the movement, and the number of individuals in the subgroup is the number of individuals that are consistent with its target within the reachable distance.

[0132] Step 3: Characterize the relationship between devices in the unmanned cluster: Calculate the attraction and repulsion between devices in the unmanned cluster.

[0133] In universal gravitation, any two particles have a force of attraction in the direction of the line connecting their centers. The magnitude of this gravitational force is proportional to the product of their masses and inversely proportional to the square of their distance, that is, Among them, M and m are the masses of the two particles, g is the acceleration of gravity, and r is the distance between the two particles.

[0134] Assuming that the repulsive force between unmanned devices is consistent with the universal gravitation, and making fine adjustments based on the universal gravitation formula, the formula for the attraction and repulsion between individuals is as follows:

[0135]

[0136]

[0137] Where K(i) and K(j) represent the degrees of unmanned device individuals i and j respectively. Assume that the initial relative position coordinates of the unmanned cluster are (0,0), and the relative position coordinates of unmanned device i at time q are (x i,q ,y i,q ),d ij is the Euclidean distance between device individuals i and j, k is a constant coefficient, and is adaptively adjusted according to the length of the unmanned device experimental scene.

[0138] Step 4: Define the local reachable density and formally calculate the size and range of the unmanned cluster to facilitate the subsequent calculation of the outlier index value and determine whether the individual can update the subgroup state through its own outlier, thereby updating the state of the unmanned cluster.

[0139] The concept of local reachable density is used to describe the size and range of unmanned cluster subgroups. The larger the reachable density value, the more unmanned devices there are in the subgroup. The local reachable density can be quantitatively evaluated according to the following definition.

[0140] Definition 1 (k-distance neighborhood)

[0141] The individual positions in the unmanned cluster are regarded as points in the plane, and point O is the position of the unmanned device, which represents the device. This describes the size and range of the subgroup based on the k-neighbor distance of each unmanned device. The k-distance neighborhood is a distance d (O,Pk(O))Describes the range of the radius, which is a circle, d (O,Pk(O)) Represents unmanned equipment individuals O and P k(o) The distance between k(O) is the set of points closest to point O). The kth distance neighborhood of point O is called N k (O), which means that the center is point O and the (O,Pk(O)) is the radius range, K k (O) indicates that N k (O) is a set of individuals within the range. For any point P in the point cluster i , S is the corresponding point cluster.

[0142]

[0143] Definition 2 (kth distance)

[0144] Assume P O It is a point in the point cluster S, which is the cluster center of the current state. The individual in S is represented by P i(O)， , according to individual to P O Distance record P i(O) ,

[0145] The kth distance is denoted as d k(O) :refers to the subgroup cluster center point P O To point P k(O) The distance between individuals in the subset and O is sorted, P k(O) is the kth closest point to O. When the number of individuals within the perception distance is less than k, then P k(O) The value of is the kth distance.

[0146] For example, Figure 2 When k = 5, d 5(O) In the diagram, at this time, with O as the center and d 5(O) The points on the circle with radius are all points at the 5th distance.

[0147] The kth distance satisfies the following conditions:

[0148] ① In the point cluster S, with O as the center, there is a set K consisting of at least k points k (O)∈S\{O},K k (O) The set of individuals is called the k-distance neighborhood.

[0149] K k (O) = {P 1(O) , P 2(O) , ...P i(O) , ...P k(O) |i≤k}, so that d (O,Pi(O)) ≤d k(O), P i(O) ∈K k (O).K k (O) Number of individuals in the set |K k (O)| may be greater than k because there are situations where the distances from point O to multiple points are equal, such as Figure 2 Medium, d 5(O) =d 6(O) , for cases where the distances are equal, the order values ​​are randomly assigned.

[0150] ② In the point cluster, there is at least a set K consisting of k-1 points k-1 (O)∈S\{O},K k-1 (O) = {P 1(O) , P 2(O) , ...P i(O) , ...P (k-1)(O) |i≤k-1}, so that d (O,Pi(O)) <d k(O) .

[0151] So far, the kth distance d of individual O k(O) , refers to the subgroup cluster center point P O To point P k(O) The distance between individuals in the subset and O is sorted, P k(O) is the kth closest point to O. Therefore, the kth distance of point O is d k(O) =d (O,Pk(O)) .

[0152] Definition 3 (reachable distance)

[0153] To calculate the local reachable density, it is necessary to express the distance from the cluster center to the surrounding individuals, which is called the reachable distance. The reachable distance is related to the parameter k of the kth distance.

[0154] Relative to a cluster center O, define point O (denoted by P o ) and P j The kth reachable distance between o k (Po,Pj) When an individual is in a circle with O as the center and the kth distance as the radius, its kth reachable distance is d k(O) , Figure 3 The circles in represent the reachable distance range of each device point when k=3.

[0155] When a device individual is outside the circle (when there are multiple individuals with equal distances to the center point, such as when the 5th and 6th individuals have the same distances to the center point, then the 5th reachable distance is the range of the circle with 5 and 6 as radii, and the 6th reachable distance is also the circle, but the points closest to the 6th also need to be considered. At this time, the point P 7Outside the circle, such as Figure 4 As shown in (d(O,P 8 ) is actually the 7th reachable distance, because this point is the 7th closest point to O), and the kth reachable distance is the actual distance from the cluster center to the individual. Its formal description is:

[0156]

[0157] θ is a parameter, set to the maximum reachable distance, when When P o and P j Unreachable. Usually, θ is taken as the maximum distance that each unmanned device can perceive without communication.

[0158] The above formula indicates that the k points closest to point O have the same k-th distance. θ ensures that the distance from the outlier to O is not too far (let θ be the maximum sensing range of the unmanned device). The set of individuals within the k-th reachable distance range centered on point O (within θ) is denoted as Kd k (O).

[0159] Definition 4 (Locally accessible density)

[0160] The local reachable density represents the density of individuals within the neighborhood of θ with point O as the center. It is used to characterize the density of individuals around point O. When the local reachable density increases, it means that the number of individuals in the current range increases, and the outlier index of individuals in the range increases, otherwise the outlier index decreases. When subgroups aggregate, the local reachable density increases, but it cannot be greater than a specified threshold to avoid collision.

[0161] The kth reachable distance determines the local reachable density, which represents the reciprocal of the average kth reachable distance of all points from point O to O within the range of θ. The kth local reachable density lrded centered on O k (O)Formally defined as:

[0162]

[0163] Among them, K k (O) represents the set of individuals within the kth reachable distance range centered on individual O. |K k (O)| represents the size of the set.

[0164] Step 5: According to the local reachable density and the attraction and repulsion between individuals, design the local outlier index to determine the state of individuals in the unmanned subgroup: determine whether to separate from the original subgroup or aggregate with other subgroups.

[0165] Definition 5 (Local Outlier Index)

[0166] The local outlier index is a measure of the tendency of an individual to separate from the original subgroup, which is determined by the local reachability density and the distance between the individual and other individuals in the subgroup. When the local outlier index increases, it means that the separation force of the individual from the original subset increases.

[0167] Based on the above definition, the calculation formula of the local outlier index can be obtained as follows.

[0168]

[0169] Among them, P i (O)∈K k (O), K k (O) represents the set of individuals within the kth reachable distance range centered on individual O.

[0170] The outlier index is used to measure the willingness of unmanned equipment to leave the original subgroup. If an unmanned equipment has a greater attraction and repulsion with other individuals outside its original subgroup, it will leave the original subgroup and enter a new subgroup; otherwise, the degree value of the unmanned equipment is adjusted to return to the original subgroup. In formula 7, f(O,j) is the attraction and repulsion of individual O in the subgroup by individual j, and α and β are artificially set weight coefficients, which can be adjusted according to the size of the unmanned cluster and the specific problems it faces. It can usually be set to α=β=0.5; lrded k (O) represents the kth local reachable density centered at O.

[0171] Figure 5 Indicates the subgroup position of point P, according to the distance O of the unmanned equipment P 1 and O 2 The position of the unmanned device P is calculated by calculating its outlier index. 2 in the same subgroup.

[0172] Step 6: Based on the outlier index in step 5, the individual determines whether to separate from the original subgroup and aggregate with the new subgroup to update the subgroup status. By updating the subgroup status and the status of the individual unmanned equipment, the overall status of the unmanned cluster can be updated.

[0173] Definition 6 (Subgroup State)

[0174] Subgroup C MN The state is represented by S(C MN ), which can be described as a triple, storing: 1) the number of individuals in the subgroup; 2) the status of each unmanned device in the current subgroup (direction, speed and task); 3) the status of the subgroup's corresponding target (direction, speed and remaining task amount). Among them, N is the number of unmanned devices in the subgroup, and M is the number of tasks that the subgroup needs to perform. Individuals with the same tasks are considered to be in the same subgroup, and the number of individuals in the subgroup is calculated based on this.

[0175] The state S(C MN ) is related to the number of individuals in the subgroup and the target completed workload of the current subgroup. The completed workload of the current subgroup is represented by the working time and the number of individuals in the subgroup.

[0176] The outlier index of individual O is calculated according to formula 7. When the outlier index is greater than the artificially specified threshold, the individual leaves the original subgroup and is attracted to the subgroup where the neighbor individual of a non-same subgroup with the greatest attraction and repulsion to it is located. At the same time, it is separated from the atomic group and aggregated with the subgroup where the neighbor individual is located. When there are no other subgroups around it, the individual is separated from the original subgroup and becomes an independent new subgroup.

[0177] When an individual separates from the original subgroup or aggregates into another subgroup, the corresponding triplet information representing the subgroup status is updated; accordingly, the subgroup information in the individual attribute is also updated. The subgroup information and individual information together represent the updated cluster status, including the number of subgroups, the number of individuals in each subgroup, and the seven-tuple information related to the speed and position of each individual.

[0178] Provide the working principle and working process of the present invention below:

[0179] The principle of goal-oriented decentralized cluster reconstruction for fragmented information is to achieve sub-group reconstruction through separation and aggregation of sub-groups when each unmanned device can only obtain limited fragmented information, realize cluster formation and target allocation, and make the cluster an intelligent device cluster with self-organization, self-adaptation, self-learning and self-collaboration.

[0180] Specifically, the unmanned cluster is regarded as a whole composed of several subgroups. Each subgroup performs a certain task separately, and the subgroups together form a cluster formation. In the process of executing the task, according to the target situation and the status of each subgroup itself, the concepts of local reachable density and separation index of each device are used to realize the separation-based reconstruction of the subgroups.

[0181] The local reachable density describes the size and range of each subgroup in the unmanned cluster. The larger the local reachable density, the more unmanned devices there are in the subgroup. At the same time, based on the local reachable density and combined with the target situation, the separation index of each device relative to the subgroup is defined to determine whether a device is separated from the atomic group and reconstructed into a new subgroup.

[0182] The principle of cluster reconstruction based on separation of unmanned subgroups is as follows: the behavior of each unmanned device in the subgroup is determined by calculating the local reachable density and separation index of the subgroup where the device is located. The larger the separation index, the greater the power of the device to leave the current subgroup. When the separation index is greater than the specified threshold, the device will leave the current subgroup and determine its new subgroup or become a separate subgroup based on the state of the target or other neighbor companions that attract it, forming a cluster reconstruction based on separation, and realizing the goal-oriented unmanned cluster adaptive formation.

[0183] The goal-oriented unmanned cluster for fragmented information is based on the idea of ​​cluster reconstruction based on separation: first, define the expression form of unmanned equipment; then, define the relationship between the equipment and its surrounding equipment to form the expression mode of the unmanned cluster; thirdly, design the concept of outlier index based on local reachable density, and combine it with the target allocation problem, from the perspective of separation of equipment and its sub-group, construct a cluster reconstruction based on separation, realize dynamic formation of clusters, and realize self-organizing, self-adaptive, self-learning and self-collaborative clusters.

[0184] The goal-oriented unmanned swarm dynamic subgroup reconstruction algorithm is described as follows:

[0185] enter:

[0186] 1. N unmanned devices and the absolute position of each device;

[0187] 2. Assume that there are M targets and the positions of each target.

[0188] Output:

[0189] M subgroups are in a relatively stable motion state and move toward the target position;

[0190] Algorithm steps:

[0191] 1) Initially, N unmanned devices are clustered based on distance to form M subgroups;

[0192] 2) Each unmanned device moves toward the target at a specified speed and direction;

[0193] 3) The speed and direction of the device in each subgroup is determined by the speed and direction of other devices in its subgroup;

[0194] 4) Calculate the local reachability density and separation index of each device in the subgroup;

[0195] 5) When the separation index of a device in a subgroup is greater than the specified threshold, the device will leave the atomic group and follow the device with the greatest attraction to form a new subgroup, or become a subgroup alone;

[0196] 6) Each subgroup reaches the target position and the algorithm ends.

[0197] The parameters involved in the algorithm and their common value settings are shown in Table 1 below:

[0198] Table 1 Parameters of the Decentralized Unmanned Cluster Simulation Experiment

[0199] Coordinates of the upper right corner of the region: (5000，5000) Unmanned equipment perception radius: 500 Unmanned equipment attack radius: 100 Unmanned equipment speed: 20 Target speed: 20 Distance between devices in a group: 20 Minimum number of devices in the group: 2 K (Kth reachable distance): 5 α (weight when finding outlier index): 0.6 α (weight when seeking the subgroup state): 0.8 Subgroup status domain value: 5 Outlier index threshold: 0.8 Reachable density threshold: 0.8

[0200] Among them,

[0201] (1) The position where the target appears is set randomly. Initially, the target position is at the top or the rightmost. Denote the coordinate range as (Xmax, Ymax). Then the range where the target appears is:

[0202] (X, Y|Xmax / 2 < X <= Xmax, Y = Ymax) or (X, Y|X <= Xmax, Ymax / 2 < Y <= Ymax)

[0203] (2) The landing point of the target is set randomly. According to the normal situation, the restriction condition is set that the ordinate is 0 and the abscissa is less than the abscissa at the initial time.

[0204] (3) Usually, the number of unmanned devices is set to be 5 times or more than the number of targets.

[0205] Set different numbers of unmanned devices and target tasks to verify whether the unmanned cluster can complete the tasks of multiple targets under the condition of fragmented information (only being able to ensure the perception of the information of surrounding individuals).

[0206] Table 2 Simulation Running Time and Task Success Rate under Different Numbers of Unmanned Devices and Targets

[0207] Serial number Number of unmanned equipment Target quantity Operation time (milliseconds) Success rate 1 80 5 347.8319 100％ 2 80 10 191.1741 100％ 3 80 20 302.8736 100％ 4 80 30 269.4441 100％ 5 100 10 344.5067 100％ 6 100 20 339.5740 100％ 7 100 30 382.0017 100％ 8 120 10 451.7124 100％ 9 120 30 544.9464 100％ 10 120 50 646.7788 100％ 11 150 30 734.5293 100％ 12 150 50 780.1226 100％ 13 200 30 1241.1948 100％ 14 200 50 1562.6077 100％

[0208] As can be seen from Table 2 above, the unmanned devices can achieve the complete execution of the targets. In terms of the running time, when the number of unmanned devices is 150 and the number of targets is 50, the algorithm time is 780.1226 milliseconds, still less than 1 second; when further increasing the number of unmanned devices and targets, the operation time is also relatively extended, but it is still acceptable.

Claims

1. A goal-oriented decentralized unmanned swarm dynamic formation method for fragmented information. It is characterized in that The method comprises: Step 1: Divide the unmanned cluster into subgroups consisting of several unmanned devices; Step 2: Obtain unmanned device description information, unmanned device subgroup description information, and the attraction and repulsion between devices in the unmanned cluster; Step 3: Obtain the local reachable density of each device in the unmanned cluster subgroup. The local reachable density represents the individual density within the θ neighborhood with the unmanned device O as the center. Point O represents a certain unmanned device and its location in the subgroup, and the θ neighborhood is the maximum self-perception distance of the unmanned device in the absence of communication. Step 4: According to the local reachable density and the attraction and repulsion between individuals, the local outlier index is obtained. The calculation formula of the local outlier index is: Among them, P i (O)∈K k (O), K k (O) represents the set of individuals within the kth reachable distance from individual O, |K k (O)| represents the size of the set; f(O,j) is the attraction and repulsion of individual O in the subgroup by individual j; α and β are artificially set weight coefficients, which can be adjusted according to the size of the unmanned cluster and the specific problems it faces; lrded k (O) represents the kth local reachable density centered at O; Represents unmanned equipment individuals O and P i (O) The minimum distance between Step 5, determine the individual behavior according to the local outlier index in step 4; when the local outlier index is greater than the specified threshold, the individual leaves the original subgroup and is attracted to the subgroup where the neighboring individual of the non-same subgroup with the greatest attraction and repulsion is located; when there are no other subgroups around it, the individual separates from the original subgroup and becomes a new subgroup independently; Update the triplet information representing the subgroup status corresponding to the unmanned device individual, and update the subgroup information in the attributes of the unmanned device individual; Step 6: By updating the status of the sub-groups and the status of the individual unmanned devices, the overall status of the unmanned cluster is updated to achieve target-oriented dynamic formation of each sub-group.

2. According to claim 1, the goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information, It is characterized in that The unmanned equipment description information includes: theoretical target number, perceived target number, unmanned equipment number, coordinates, speed, direction and number of individuals in the subgroup; The unmanned device subgroup description information includes: the unmanned device numbers contained in the subgroup, the virtual edges connecting the devices in the unmanned subgroup, and the degree-value function related to degree in the unmanned subgroup network.

3. According to claim 2, the goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information, It is characterized in that The function of degree-degree correlation in the unmanned subgroup network is: Among them, d ij represents the distance between unmanned equipment individual i and the individual j closest to i, ρ σ (i) represents the local reachability density of individual i in the neighborhood of σ, which is used to describe the number of individuals in the subgroup, θ 1 and θ 2 is the coefficient, θ 1 +θ 2 =1, σ is a constant parameter, which can be set according to the characteristics of the unmanned cluster and is used to describe the range that the unmanned equipment can perceive on its own.

4. According to claim 3, the goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information, It is characterized in that The calculation formula of the attraction and repulsion between the devices in the unmanned cluster is: Among them, K(i) and K(j) represent the degrees of unmanned equipment individuals i and j respectively. Assume that the initial relative position coordinates of the unmanned cluster are (0,0), and the relative position coordinates of unmanned equipment i at time q are (x i,q ,y i,q ),d ij is the Euclidean distance between device individuals i and j, k is a constant coefficient, and is adaptively adjusted according to the length of the unmanned device experimental scene.

5. According to claim 4, the goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information, It is characterized in that The kth local reachable density lrded centered at O k The formula for (O) is: Among them, K k (O) represents the set of individuals within the kth reachable distance from individual O, |K k (O)| represents the size of the set; the formula for the kth reachable distance is: Among them, θ is a parameter, which is set to the maximum reachable distance. When P is unmanned, it indicates o and P j Unreachable; d k(O) Subgroup cluster center point P O To point P k(O) The distance between individuals in the subset and O is sorted, P k(O) is the kth closest point to O; d (O,Pj(O)) Represents unmanned equipment individuals O and P j The distance between j is the j-th closest point to point O.

6. The goal-oriented decentralized unmanned swarm dynamic formation method for fragmented information according to claim 5, It is characterized in that The value of θ is the maximum distance that each unmanned device can perceive itself without communication.

7. The goal-oriented decentralized unmanned cluster dynamic formation method for fragmented information according to claim 1, It is characterized in that In step 5, the triple information is S(C MN ), including: Subgroup C MN The state is represented by S(C MN ), which can be described as a triple, which stores the number of individuals in the subgroup, the status of each unmanned device in the current subgroup, and the status of the subgroup's corresponding target, where N is the number of unmanned devices in the subgroup, and M is the number of tasks that the subgroup needs to perform. Individuals with the same tasks are considered to be in the same subgroup, and the number of individuals in the subgroup is calculated in this way.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor runs the computer program stored in the memory, the steps of the method according to any one of claims 1 to 7 are performed.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a plurality of computer instructions, and the plurality of computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

10. A computer program product, It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images