CN113316118B

CN113316118B - Unmanned aerial vehicle cluster network self-organizing system and method based on task cognition

Info

Publication number: CN113316118B
Application number: CN202110600519.3A
Authority: CN
Inventors: 尹栋; 李�杰; 贾圣德; 相晓嘉; 王祥科; 喻煌超
Original assignee: National University of Defense Technology
Current assignee: National University of Defense Technology
Priority date: 2021-05-31
Filing date: 2021-05-31
Publication date: 2022-07-19
Anticipated expiration: 2041-05-31
Also published as: CN113316118A

Abstract

The invention discloses an unmanned aerial vehicle cluster network self-organizing system and a method based on task cognition, wherein the system comprises the following components: the application layer is used for establishing a representation method and a model from the cluster task domain to the network information domain to obtain a cluster internal information cross-linking relation; the network layer is used for constructing a network logic topological relation based on the information association relation of each node in the cluster established by the application layer and generating a network topological relation graph; the link layer is used for realizing network structural design and generation, constructing a network link through a network logical topological relation constructed based on the network layer and reconstructing a dynamic network link when network dynamic change is generated; and the physical layer is used for constructing a simulation environment so as to perform simulation test on the performance of the cluster network. The invention can obviously improve the searching and delivering efficiency, the information dimension and the control precision of a large-scale unmanned aerial vehicle system, improve the reliability and the efficiency of an unmanned intelligent group system, and flexibly cope with complex environments and emergencies.

Description

Unmanned aerial vehicle cluster network self-organizing system and method based on task cognition

Technical Field

The invention relates to the technical field of unmanned aerial vehicle cluster control, in particular to an unmanned aerial vehicle cluster network self-organizing system and method based on task cognition.

Background

The unmanned aerial vehicle cluster is a distributed system which comprehensively integrates a large number of unmanned aerial vehicles under an open system architecture, takes cooperative control among platforms as a basis and aims at improving cooperative task capability. Unmanned aerial vehicle systems are gradually developing from 'large and full' single-platform high autonomy to 'small and fine' low-cost micro-miniature group intelligence, and the development of micro-miniature unmanned aerial vehicle clusters has been paid high attention. The basis of unmanned aerial vehicle cluster application lies in network communication, and due to the restriction of volume, weight, energy and software and hardware technical levels, the autonomous level of the unmanned aerial vehicle is generally not high (less than or equal to 5), so that a cluster needs to perform environment perception, decision planning, behavior synchronization and the like through a large amount of information interaction in the process of executing a dynamic task in a complex environment, and the individual deficiency is made up by group advantages. The key points for improving the use efficiency of the cluster are to construct a network architecture of group task behavior characteristics and research a network communication technology of 'emergency response and rapid deployment'. However, for the unmanned aerial vehicle cluster network architecture which has been developed and applied, a design from the bottom to the top is mostly adopted, a network is firstly constructed based on the existing communication equipment and communication mechanism, optimization is performed in the information assurance process, and planning and designing are not performed from the top-level task requirement, so that the supply-demand relationship between the network architecture and the tasks is inverted at the end.

The unmanned aerial vehicle cluster application has the characteristics of large quantity, wide range, high speed, flexibility, frequent change of space-time relationship, changeable tasks, cross-regional scheduling and the like, and provides great challenges for defining the aspects of information transmission, network structure, dynamic optimization and the like among unmanned aerial vehicles, and the unmanned aerial vehicle cluster application mainly has the following problems:

1. in terms of a network layer, a large-scale unmanned aerial vehicle cluster network is a typical complex network, and a network generation method is designed by starting from typical task characteristics of a cluster and establishing an individual behavior and an information transmission model between individuals. At present, the task process of the unmanned aerial vehicle is generally abstracted into a particle behavior model with certain probability distribution characteristics, and various information of concurrent interaction between nodes is uniformly expressed into a transmission capacity value between the nodes, so that the network architecture theory research is separated from the cognition on the cluster task.

2. In the aspect of a link communication layer, research in the prior art focuses on communication waveform design, channel design and access technology, but does not consider information service types and transfer characteristics in a cluster task flow, does not consider associated transfer mapping and characterization from a cluster task domain to an information domain, and cannot guide the design and construction of a cluster network architecture from the requirement of the task flow on information.

3. The contradiction between loose cluster organization and tight intra-cluster communication makes the generation of cluster networks challenging. On one hand, the unmanned aerial vehicle cluster has the characteristic of taking tasks/subtasks as the center, each unmanned aerial vehicle can be scheduled and distributed on line, the characteristics of flexible network access and exit of nodes and rapid fusion and separation of sub-networks are formed, and the node organization structure is in a loose coupling form. On the other hand, in the task flow, information interaction between nodes surrounding the same subtask is frequent, information transmission between subnets in the cooperative task is tight, each link of the task is closely coupled with information quality, and meanwhile, the movement and the state of each node participating in the task influence the stability and the transmission quality of a data link, so that task participants are tightly coupled with information transmission. The loose task organization and behavior of the clusters bring great challenges to the close information interaction between nodes in task execution. In the prior art, a network generation method cannot determine a coupling mechanism between task behaviors and information services according to the group behavior characteristics of the unmanned aerial vehicles, so that the mutual coupling characteristics of a link layer, a network layer, a transmission layer and an application layer cannot be matched to realize network generation.

4. The unmanned aerial vehicle gives consideration to tasks and communication, contradictions exist between task re-planning and network optimization in a complex dynamic scene, and network robustness and transmission stability face challenges. In a complex task scene, such as low-altitude close-range reconnaissance monitoring, earthquake-resistant disaster relief emergency communication and the like, tasks are possibly re-planned (task re-allocation, airway re-planning and the like) along with situation development, and the organization relationship of a cluster is changed; meanwhile, a few unmanned aerial vehicles possibly lose control or are destroyed in a task, so that the network topology changes, links need to be reconstructed, the network topology needs to be adjusted, the routing needs to be optimized, and how to ensure continuous and stable network information transmission in the dynamic change process is an optimization problem. However, each drone in the cluster is both a task performer and a network participant, and the behavior of the drone is planned and constrained in two different dimensions from network optimization and task performance, so that the solution cannot be achieved. In the prior art, the cluster control method does not consider the role distribution and role change of each node in the cluster in the task and the task relevance of the node time changing along with space, and cannot truly represent the dynamic evolution process in the cluster control to realize dynamic and flexible control.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: aiming at the technical problems in the prior art, the invention provides the unmanned aerial vehicle cluster network self-organizing system and method based on task cognition, which are simple in structure, high in reliability and efficiency and strong in flexibility.

In order to solve the technical problems, the technical scheme provided by the invention is as follows:

an unmanned aerial vehicle cluster network self-organizing system based on task cognition comprises:

the application layer is used for establishing a representation method and a model from the cluster task domain to the network information domain to obtain a cluster internal information cross-linking relation;

the network layer is used for constructing a network logic topological relation based on the information association relation of each node in the cluster established by the application layer and generating a network topological relation graph;

the link layer is used for realizing network structural design and generation, constructing a network link through a network logical topological relation constructed based on the network layer and reconstructing a dynamic network link when network dynamic change is generated;

and the physical layer is used for constructing a simulation environment so as to perform simulation test on the performance of the cluster network.

Further, the application layer includes:

the task flow representation module is used for characterizing the cluster task based on the task information flow and the node information flow;

And the topological relation graph generating module is used for establishing an information transfer relation between a cluster task information flow and each unmanned aerial vehicle node based on the cluster task represented by the task flow representation module, and forming an information association graph g of each node in the cluster so as to construct the network logic topological relation.

Further, the network layer includes a network topology generation module, configured to construct the network logical topology relationship according to the information association relationship of each node in the cluster; the network topology generation module is a method for discovering communities based on conflict graph transformation, different information services in task information flow are used as requirements of different service communities, node information flow is used as a node logic cross-linking relation in the community, community division is carried out according to logic connection of information transmission, and a community distribution set C is formed_mAnd generating the network initial topological graph.

Further, the link layer includes:

the link construction module is used for constructing and forming a two-layer clustering network structure with network nodes in a layering way through direct link or cooperative communication through relay nodes, and constructing a ground access network to instantly access the ground network in a zoning way to form a three-layer clustering network structure;

The hierarchical routing module is used for constructing a routing mode of communication among nodes based on a three-level clustering cluster network structure formed by hierarchical construction of the link construction module;

and the link reconstruction module is used for searching potential connectable nodes by taking the maximum node power as a search radius when network dynamic change is generated, screening and filtering the nodes by taking preset connectable time and preset connectable degree as constraint conditions, then iterating to obtain an optimal node, and performing link connection to finish network reconstruction.

Further, the method comprises the following steps:

s1, establishing an information transfer relation between a cluster task information stream and each unmanned aerial vehicle node according to a cluster task at an application layer to form an information association graph of each node in a cluster;

s2, at the network layer, establishing a network logic topological relation based on the information incidence relation of each node in the cluster established by the application layer, and generating a network topological relation graph;

s3, constructing a network link based on the network logic topological relation graph at the link layer;

s4, when network dynamic change occurs, real-time reconstruction of a network link is carried out on the link layer;

and S5, establishing a simulation environment in the physical layer, and performing simulation test.

Further, in the step S1, the cluster task is characterized based on the information flow dynamic hypergraph model to form an information association graph of each node in the cluster, and the specific steps include:

s101, cluster task space construction: adopting a hierarchical decomposition type structured modeling method to subdivide the cluster tasks into meta tasks layer by layer according to the top-down sequence, wherein the meta tasks are relatively independent and directly executable tasks and map the analyzed objects in a hierarchical structure form;

s102, constructing a cluster task information flow dynamic hypergraph model: representing nodes of the unmanned aerial vehicle as vertexes of a hypergraph, abstracting the meta-task into a hyper-edge, and connecting vertexes associated with the hyper-edge according to an association matrix between the meta-task and the nodes of the unmanned aerial vehicle; in each hyper-edge, associating the associated vertexes by using a connecting line according to the content and the node type of the meta-task to obtain a static hyper-graph model; and finally, arranging the static hypergraph models at multiple moments according to a task time sequence to obtain a dynamic time-varying UIF hypergraph model in the whole task process so as to represent the information association relation between nodes and obtain the information association diagram of the nodes.

Further, in the step S2, a network topology relationship diagram is generated by adopting a method of discovering communities based on conflict diagram transformation, wherein after an information association diagram between network nodes of a node information stream in a cluster network is obtained, a conflict diagram mode is adopted, neighborhood influence in n hops of each node is considered, and a mutual influence relationship between the node and an edge is judged; in the process of carrying out community discovery on a conflict graph influenced by n hops, a group of discovered communities comprises different node subsets, wherein each node and at most n-hop neighbors thereof are located in the same subset, and the influence degree information is utilized to identify hierarchical community distribution with larger or smaller scales at different levels;

the step of generating the network topological relation graph by adopting the conflict graph transformation discovery community method comprises the following steps:

s201, obtaining a network graph with v nodes and epsilon links

N-hop conflict graph of

Identifying the n-hop conflict graph

Of (5) to get conflict graph community C'_m

S202, community C 'to the conflict graph'_mCarrying out reverse transformation to transform each node into an edge to obtain the network graph

A new overlapping community, namely, a part of nodes exist in a plurality of communities;

S203, finding out target nodes belonging to a plurality of communities, and keeping the target nodes in a community set C with the target nodes having the maximum link number to obtain a final community distribution set after the overlapped nodes are deleted finally;

s204, the final community distribution set and the community distribution C obtained in advance_MComparing, and regenerating a community distribution set according to a comparison result until an optimal community set C is obtained;

and S205, outputting a network topology relation graph with the optimal community set C in a topology graph mode.

Further, in step S3, a heuristic link construction method based on the combination of average flow capacity ANFC enhancement and cooperative communication is used to form a global link connection diagram of the cluster network to generate a cluster initialization network link, and the specific steps include:

s301, acquiring network initial conditions:

to be provided with

A node, an epsilon link and a link capacity of

Initial network of, network

The number of nodes of the link which need to be added is k;

s302, traversing from 1 to N nodes, searching for a node with a degree of p, storing the node in a node _ set, and storing a node with a degree of k and a preset r hop distance in the node _ set;

s303, after the searching is finished, judging whether the node _ set has the condition of meeting the k link establishment, if so, returning to the step S302 for the next round of traversal searching, and turning to the step S304; if not, go to step S305; (ii) a

S304, calculating Euclidean distances among all possible node pairs in the node _ set, arranging the node pairs according to the distance values in a descending order, adding k links to the first k nodes in the node _ set, and adopting variable capacity C_varAnd a fixed capacity C_fixAfter each link is assigned, adding the direct link graph between the nodes

S305, introducing a cooperative communication relay node, carrying out link establishment and network supplement communication, assigning values to link capacity, and forming a relay graph

Further, in step S3, the communication between the nodes uses a look-ahead routing algorithm of hierarchical clustering, that is, a neighbor node set is constructed based on a range of n hops to obtain a neighborhood, and destination node information is searched using the degree of a node as the breadth and the n hops as the depth, where the look-ahead routing algorithm of specific hierarchical clustering specifically includes: searching neighbor nodes in n hops, and if a target node is in the neighborhood of a source node, directly sending a message to the target node; if not, the source node selects a neighbor as an intermediate forwarding node according to the position of the destination node, and forwards the message to the intermediate forwarding node; and after receiving the message, the intermediate forwarding node judges whether the target node is in the neighbor domain, if so, the intermediate forwarding node sends the message to the target node, otherwise, the intermediate forwarding node continues to select the next intermediate forwarding node and forwards the message until the nodes in the community and the adjacent community are traversed.

Further, in step S4, link quality between nodes is settled in real time by using link influence entropy lile as criterion of network stability and using the constructed wireless communication model, and based on the network link map

Evaluating the stability of the network, and triggering the reconstruction of the network link when the stability is reduced to meet the preset conditions;

in the step S4, performing network link reconfiguration by using a heuristic bee colony algorithm method includes:

s401, when the local communication delay of the network rises to a preset threshold value within a specified time length so as to cause node failure in the network topology, the local communication delay of the network is added to the related node p_iThe nodes of the link can be established in the initial search area, and the population scale m and the maximum iteration number L are established_maxInitialization traversal number l_sSetting a global variable v, establishing a set of nodes to be reconstructed of the unmanned aerial vehicles with the residual numbers at one time, and guiding all the nodes to complete traversal;

s402, randomly generating initial solutions, recording the adaptive value of each solution, and selecting a neighborhood node with the highest fitness as an optimal solution after neighborhood optimization;

s403, the local optimization times and a preset threshold value L are set_maxBy comparison, if l_s＞L_maxGiving up the current solution, randomly extracting new nodes from the rest bee colony to solve and update, and juxtaposing traversal times l _s0; if the global variable v is P _ num, and P _ num is the number of nodes involved in link reconstruction, the optimization solution is endedOtherwise, jumping to step S402;

and S404, outputting the finally obtained optimal solution, wherein the corresponding link connection is the optimal link selection of the current magnetic reconstruction.

Compared with the prior art, the invention has the advantages that:

1. according to the invention, an intelligent group network construction and maintenance system of the unmanned aerial vehicle cluster is formed from the perspective of task planning and network control by taking unmanned aerial vehicle group application tasks as requirements, the unmanned aerial vehicle cluster is regarded as an intelligent group, the task targets are correlated, task related information interaction is used as interconnection and intercommunication to form a network, a network structure of multi-level and overlapped communities is constructed, group behaviors, task intentions, information interaction and a network structure generation process can be organically combined with each other, the search and delivery efficiency, the information dimension and the control precision of a large-scale unmanned aerial vehicle system can be remarkably improved, the reliability and the efficiency of the unmanned intelligent group system are improved, and the unmanned intelligent group system can flexibly cope with complex environments and cope with emergencies.

2. The unmanned aerial vehicle cluster is regarded as an intelligent cluster, a network is formed by interconnecting and intercommunicating task targets and task related information, a small-world network model facing timely group communication and multidimensional characteristic quantity are established based on a cluster task flow, the community is established by information supply and demand relation, a multi-level and overlapped network structure of the community is established by a task time sequence flow, and the group behaviors, the task intentions, the information interaction and the network structure generation process can be organically combined.

3. The invention further considers a task and information coupling mechanism, constructs a cluster information flow-to-dynamic hypergraph model of a task scene by using a representation method of a group task time sequence flow and a continuous process of cooperative behaviors among individuals based on a graph theory method of an information flow model and a complex network, generates a network topology structure based on cluster characteristics and information transmission requirements, and can improve the cognition of an unmanned platform cluster system so as to form an intelligent group system.

4. The invention further establishes a multi-level mutual coordination method and mechanism of link dynamic reconstruction, network structure recombination and route regeneration in consideration of the overall stability of the network, and solves the problem of mutual disjointing in the traditional network layer and link layer dynamic adjustment.

Drawings

Fig. 1 is a schematic structural principle diagram of the unmanned aerial vehicle cluster network self-organizing system based on task awareness according to this embodiment.

Fig. 2 is a schematic flowchart of the task awareness-based unmanned aerial vehicle cluster network self-organization implemented by this embodiment.

Fig. 3 is a schematic diagram of the information flow of the cluster task in a specific application embodiment.

Fig. 4 is a schematic diagram of the construction of a conflict graph from a given network base topology in a specific application embodiment.

Fig. 5 is a schematic diagram of a two-layer clustering structure of the cluster air network constructed in this embodiment.

Fig. 6 is a schematic diagram of a three-level cluster network structure constructed in the present embodiment.

Detailed Description

The invention is further described below with reference to the drawings and specific preferred embodiments of the description, without thereby limiting the scope of protection of the invention.

As shown in fig. 1, the task awareness-based unmanned aerial vehicle cluster network self-organizing system of the present embodiment includes:

the network layer is used for constructing a network logical topological relation based on the information association relation of each node in the cluster established by the application layer and generating a network topological relation graph;

the link layer is used for realizing the structured design and generation of a network, constructing a network link through a network logical topological relation constructed based on the network layer and reconstructing a dynamic network link when the dynamic change of the network is generated;

According to the embodiment, the unmanned aerial vehicle cluster intelligent cluster network construction and maintenance system is formed from the perspective of task planning and network control by taking unmanned aerial vehicle cluster application tasks as requirements, the unmanned aerial vehicle cluster is regarded as an intelligent cluster, the task targets are correlated with each other, task related information interaction is used as interconnection and intercommunication to form a network, a multi-level and overlapped network structure of the community is constructed, the cluster behaviors, the task intentions, the information interaction and the network structure generation process can be organically combined with each other, the searching and delivering efficiency, the information dimension and the management and control precision of a large-scale unmanned aerial vehicle system can be remarkably improved, the reliability and the efficiency of the unmanned intelligent cluster system are improved, the complex environment can be flexibly coped with, and emergencies can be coped with.

The application layer specifically includes:

the task flow representation module is used for characterizing a cluster task based on a task information flow and a node information flow;

and the topological relation graph generating module is used for establishing an information transmission relation between a cluster task information flow and each unmanned aerial vehicle node based on the cluster task represented by the task flow representation module, and forming an information association graph of each node in the cluster so as to construct a network logic topological relation.

Because the cluster tasks have multiple types, the single-machine independent task and the multi-machine cooperative task are mutually interwoven and can be executed in parallel, and the information service requirement changes along with the progress of the task time, the task decomposition granularity is tightly coupled with the information flow description, and the cluster tasks are difficult to accurately decompose and establish a dynamic information flow model between nodes. In this embodiment, a description method for information flow is used for reference at an application layer, and a characterization method and a model from a cluster task domain to a network information domain are established to obtain a cross-linking relationship of cluster internal information, which is convenient for subsequently building a network logical topology relationship by combining the cross-linking relationship of the cluster internal information.

The network layer specifically comprises a network topology generation module, which is used for constructing a network logic topology relation according to the information association relation of each node in the cluster; the network topology generation module finds a community method based on conflict graph transformation, takes different information services in a task information stream as requirements of different service communities, takes a node information stream as a node logic cross-linking relation in the community, and transmits logic according to information Connecting to divide communities to form a community distribution set C_mAnd generating a network initial topological graph.

The link layer specifically includes:

the link construction module is used for constructing and forming a two-layer clustering network structure in a layering way, wherein the two-layer clustering network structure is formed by the cooperative communication between the network nodes through direct links or through relay nodes, and constructing a ground access network so as to instantly access the ground network in regions to form a three-layer clustering network structure;

the hierarchical routing module is used for constructing a three-level clustering cluster network structure formed by hierarchical construction based on the link construction module and constructing a routing mode of communication among nodes;

and the link reconstruction module is used for searching a potential connectable node by taking the maximum power of the node as a search radius when network dynamic change is generated, screening and filtering the node by taking preset connectable time and preset connectable degree as constraint conditions, then iterating to obtain an optimal node, and performing link connection to complete network reconstruction.

Because the cluster task process has the characteristics of wide area distribution and local area concentration, and the task has higher requirement on information timeliness, the embodiment constructs a network logic topological relation according to the information incidence relation of each node in the cluster, generates a network initial topological graph based on a conflict graph transformation discovery community method, constructs a layered clustering network structure at a link layer, and can construct a network structure which simultaneously ensures the instant transmission of local area and wide area information under the condition that the direct communication distance between unmanned aerial vehicles is limited; meanwhile, the link reconfiguration module completes network reconfiguration when network dynamic change is generated, so that in the face of node damage or topology change caused by an emergency in a complex environment, the link can be dynamically and quickly reconfigured to ensure the stability of the network.

The four-layer structure of the unmanned aerial vehicle cluster network self-organizing system is specifically as follows:

the first layer is an application layer, mainly solves the cognitive problem from the cluster task domain to the information domain, and establishes a characterization method and a model from the cluster task domain to the network information domain by referring to a description method of information flow to obtain a cluster internal information cross-linking relation.

The second layer and the third layer are respectively a network layer and a link layer, which mainly solve the problems of network generation and evolution adapting to the cluster task characteristics, construct a network logical topological relation based on the information incidence relation of each node in the cluster established by the first layer, form a network initial topological graph, simultaneously consider the link physical characteristics, realize the network link establishment by the idea of combining multi-hop and relay, and establish a dynamic network link reconstruction method to cope with the network topological mutation in a complex environment.

The fourth layer is a physical layer and mainly solves the problems of integration and verification of key technologies, specifically, a simulation environment is constructed based on a cloud computing platform and a local area network, all algorithm modules are integrated through a bus, and various performance indexes of a cluster network are tested by taking a typical cluster task as a drive. Specifically, a plurality of virtual machines are used in a physical layer to form a virtual machine cluster, each virtual machine corresponds to a simulation node of an unmanned system, a programmable switch based on FPGA (specifically, NetFPGA can be adopted) is used as a core switch, the high-speed throughput, delay switching, embedded real-time computing and other capability characteristics of the FPGA are utilized, the programmable FPGA is specially used for simulating routing equipment and real-time simulation of communication quality between nodes, two-in-one of transmission of data packets and simulation of a communication network is realized, the computation time of simulation is effectively saved, the simulation efficiency is greatly improved, the simulation efficiency can be greatly improved, and meanwhile, the simulation can be closer to the real situation.

The method for realizing the self-organization of the unmanned aerial vehicle cluster network based on the system structure comprises the following steps:

s2, at a network layer, constructing a network logic topological relation based on the information association relation of each node in the cluster established by the application layer, and generating a network topological relation graph;

s3, constructing a network link based on a network logic topological relation graph at a link layer;

s4, when the network dynamic change is generated, carrying out real-time reconstruction on a network link at a link layer;

and S5, establishing a simulation environment in the physical layer, and carrying out simulation test.

As shown in fig. 2, in the specific application embodiment, first, (the implementation step:) a cluster task pre-planning result is analyzed, an information transfer relationship between a cluster task information stream and each unmanned aerial vehicle node is established based on an information stream description method, and an information association graph of each node in a cluster is formed

And is used as a basic basis for constructing a network topological relation. Meanwhile, a wireless communication model is constructed and used as a real-time measuring and calculating basis for the link quality. (implementation step three) by using community discovery thought in the small world network, taking different information services in the task information flow as the requirements of different service communities, taking the node information flow as the node logic cross-linking relation in the community, and carrying out community division on the logic connection layer of information transmission to form a community distribution set C _mAnd network connectivity graphs

(the implementation step IV) on the basis of the logical connection graph of the information incidence relation, considering the actual physical link communication situation, adopting a heuristic link construction method based on the small world characteristics to construct links from two aspects of direct link among nodes or cooperative communication through relay nodes, and generating a hierarchical clustering network link graph

(the fifth implementation step) in the network structure chart, by using the thought of greedy routing algorithm for reference, through designing the cross-layer search strategy and effectively controlling the search breadth and depth, the distributed routing method suitable for self-organizing multi-hop transmission is providedAnd convenient and efficient online route generation is realized. Then, (implement step (c)) in order to deal with the cluster and deal with various emergencies and produce the dynamic change of the network in the course of the task, study the real-time method of restructuring of the network link, in order to maintain the stability of the network; and finally, (implementing step (c)) constructing a simulation environment and carrying out simulation test.

The unmanned aerial vehicle cluster completes tasks through cooperation among individuals with certain autonomy, and task preplanning is the premise for overall cluster actions. The mission planning result mainly comprises a subtask time sequence and logic association, the assignment of each unmanned aerial vehicle mission target and navigation paths (such as time, position, maneuvering action and the like), load use methods and rules of airborne sensors, weapons and the like. In this embodiment, a task information flow dynamic hypergraph model is specifically adopted at an application layer to construct an expression method from a task planning result to an information service requirement: introducing a hierarchical decomposition type structured modeling method, constructing an unmanned aerial vehicle cluster task space, describing basic element task information flow in a mode of a mold body, constructing an inter-node information flow dynamic hypergraph of different information services in a task process, constructing a cluster task information flow graph and a node information flow graph, and generating a node information association graph in a cluster.

In step S1 of this embodiment, a cluster task is characterized based on an information flow dynamic hypergraph model to form an information association graph of each node in a cluster, and the specific steps include:

s101, cluster task space construction: adopting a hierarchical decomposition type structured modeling method, and according to the sequence from top to bottom, the cluster tasks are subdivided into meta tasks layer by layer, the meta tasks are relatively independent and directly executable tasks, and the analyzed objects are mapped in a hierarchical structure mode;

s102, constructing a cluster task information flow dynamic hypergraph model: representing the vertex of the hypergraph by the nodes of the unmanned aerial vehicle, abstracting the element task into a hypergraph, and connecting the vertex associated with the hypergraph according to the incidence matrix between the element task and the nodes of the unmanned aerial vehicle; in each hyper-edge, associating the associated vertexes by using a connecting line according to the content and the node type of the meta-task to obtain a static hyper-graph model; and finally, arranging the static hypergraph models at multiple moments according to a task time sequence to obtain a dynamic time-varying UIF hypergraph model in the whole task process to represent the information association relation between the nodes to obtain the information association diagram of the nodes.

In the embodiment, the cluster tasks are characterized based on the information flow dynamic hypergraph model to form the information association graph of each node in the cluster, compared with the traditional method, the association transfer mapping and characterization from the cluster task domain to the information domain can be realized, and the design and construction of a cluster network architecture can be guided by the information requirement of a task flow.

In the step S101 of constructing the cluster task space, as shown in fig. 3, a hierarchical decomposition type structured modeling method is adopted, a process of refining the complex objects layer by layer is performed according to a top-down sequence, and the analyzed objects are mapped in a hierarchical structure form, and for a definite cluster task, the cluster task can be divided into a total task, a subtask and a meta task according to a difference of hierarchical decomposition granularity. The overall task is the task understanding of the cluster in the global and macroscopic level, the subtasks are intermediate nodes of hierarchical decomposition, describe the task completed by the cooperation of local and multiple computers, and can be decomposed into other subtasks or directly into element tasks according to the requirements; the meta-task is the end of the hierarchical decomposition, a relatively independent task that can be directly executed by each drone node.

Let the cluster task be T, when the decomposition granularity G ═ alpha, the ith sub-task is

When the sub-task is decomposed into granularities, the jth meta-task of the sub-task is

The drone cluster task space can be represented as (in tree structure):

wherein Ψ:

and m is the upper hierarchical aggregation function of the bottom layer task, and n is the number of the subtasks and the number of the element tasks decomposed by the ith subtask.

For each meta task

The method can be further refined into the components of factors such as sequence numbers, contents, scales, requirements, execution nodes and the like, and the hierarchical model is expressed as follows:

wherein, Ω represents a combined relation function, num represents the execution sequence of the meta-tasks, con represents the specific actions or activities for completing the meta-tasks, sca represents the measurement (time, quantity, area, etc.) of the execution degree of the meta-tasks, req represents the degree and standard which should be achieved by the meta-tasks, and the degree and standard are determined according to different task environments and targets; nod denotes a node participating in the execution of the meta-task. According to task purpose and information service type, meta task

Can be divided into 4 categories: information and intelligence type task

Finger-controlled task

Operation execution class task

Transportation task

Satisfy the requirement of

For different types of cluster meta-tasks, it can be expressed as shown in table 1 below. Thus, the cluster overall task is subdivided into a set of meta-tasks.

TABLE 1 description of Cluster Meta-task

Serial number	Content providing method and apparatus	Dimension	Require to make a request for	Executing node
					1	Search tracking	Minute (min)	Within rated time	Nodes 1-3
2	Positioning striking	Number of	Number of search tracking targets	Node 4-9
					...	...	...	...	...
n	Transporting articles	Metropolitan area range	20kmx20km	Node m-n

The detailed process of constructing the dynamic hypergraph model of the cluster task information flow in the step S102 is as follows:

Because the capability of a single micro unmanned aerial vehicle is limited, a multi-machine cooperation mode is generally adopted to complete the subtasks of the cluster, so that a 'group' with frequent information interaction is formed, and the embodiment adopts a mode of a frequent subgraph-mode body of information in a complex network to express. An Information Flow Model (IFM) is a basic unit for forming various Information flows, has a specific spatial configuration and functional attributes, and is a key structure for a cluster task from micro-motion to macro-effect, for example, the Information Flow Model in the Information affects the Information Flow action efficiency such as Information guarantee, command control, cooperative attack, and the like.

After hierarchical decomposition, the meta-tasks can be generally regarded as a specific action program chain, and the conversion of the meta-tasks causes structural changes of the action program chain. The mobility procedure chain has different complexity, the simplest is only 1 flow motif, which contains 2 nodes. Dissimilar chains of action may contain the same nodes and form "syncs" in the context of certain tasks. Let time T metatask T_iHaving k information flow patterns IFM_pWhere p ∈ [1, k ]]Then the information flow model t corresponding to Ti can be expressed as

The unmanned aerial vehicle nodes in the cluster represent the vertexes of the hypergraph, the meta-tasks represent the hyper-edges, the information flow motif represents the connecting line between the vertexes, and then the IF can be expressed as IF ═ { N ═ N₁，N_c，...，N_s，IFM₁，IFM_c，...，IFM_s}. The task Flow is expressed as a dynamic UIF hypergraph by using semantic mapping of a hypergraph model shown in a table 2 and using an Information Flow association diagram (UAV Swarm Information Flow, UIF) between nodes in a cluster.

TABLE 2 UIF hypergraph model

In a specific application embodiment, when the UIF hypergraph model is constructed, firstly, a relatively independent and directly executable meta-task set is obtained after hierarchical decomposition is performed according to task types in a total task, unmanned aerial vehicle nodes are classified and abstracted into 4 types of fixed points of the hypergraph, and different symbols are represented. And secondly, selecting a certain task moment, abstracting the meta task into a super edge, and connecting the related vertexes by using the super edge according to the meta task-unmanned aerial vehicle node incidence matrix. And then, in each hyper-edge, according to the meta-task content and the node type and the structure and the rule of the information flow motif, associating the related vertexes by using connecting lines to obtain a static hyper-graph model. And finally, arranging the static hypergraph models at a plurality of moments according to the task time sequence to obtain a dynamic time-varying UIF hypergraph model in the whole task process.

In this embodiment, the dynamic time sequence hypergraph model is used to describe a cluster task information circulation process, establish an information association relationship between unmanned aerial vehicle nodes in a cluster, generate a node association graph from a task information demand layer, and lay a foundation for community discovery in network establishment.

In the unmanned aerial vehicle cluster, local aggregation is formed by subtasks and task areas, a subset taking the tasks as a center is formed, and information interaction of all unmanned aerial vehicles in the subset is frequent and information types are multiple, so that the tasks in the areas are ensured to be completed; and the information interaction quantity among the subsets is small, the frequency is low, and the cluster command control requirement and the resource scheduling management are mainly ensured. Therefore, after the network information cross-linking relation is formed, the network needs to be divided into multiple layers and multiple subnets according to the cluster information transmission characteristics, so as to generate a network topology structure suitable for multi-service concurrent transmission. In the embodiment, a mode of community discovery in a complex network is adopted to construct a node set with similar tasks, behaviors and information service attributes in a cluster network, and the problem of constructing a hierarchical clustering structure of the cluster network according to information service requirements is solved by determining layer subnet division by utilizing community distribution.

According to the embodiment, a multi-level multi-subnetwork structure is constructed by adopting a community discovery method based on conflict graph transformation according to the relation between information timeliness and transmission delay in a hop avoidance mechanism. In many community discovery strategies, the most critical node or edge is generally judged based on a selected parameter (such as betweenness centrality of the node or edge), global network information is often required to be acquired, so that all nodes or edges of a network need to be traversed, and the time complexity of a calculation process is high. In this embodiment, on the basis of defining the information association graph among the network nodes of the node information flow in the cluster network, in step S2, a network topology relationship graph is generated by using a method for discovering a community based on a collision graph transformation, that is, the mutual influence relationship between the nodes and the edges is determined by taking the influence of the neighborhood within n hops of each node into account. For example, fig. 4 shows a process of constructing a conflict graph from a given network basic topology, where fig. 4a) is a topology graph of node-link transformation, and fig. 4b) to 4d) are in the form of conflict graphs when 1, 2, and 3-hop links exist between two nodes, respectively. In the n-hop influenced collision graph for community discovery process described above, the discovered set of communities will contain different subsets of nodes, where each node and its at most n-hop neighbors will be in the same subset. Meanwhile, the influence degree information is used for identifying the hierarchical community distribution with larger or smaller scale in different layers.

The specific steps of generating the network topology relationship diagram by adopting the conflict diagram transformation-based community discovery method in this embodiment include:

s201, obtaining a network graph with v nodes and epsilon links

N-hop conflict graph of

Identifying n-hop conflict graphs

Of (5) to get conflict graph community C'_m；

S202, community C 'of conflict graph'_mCarrying out inverse transformation to transform each node into an edge to obtain a network graph

s205, outputting a network topology relation graph with the optimal community set C in a topology graph mode;

in a specific application embodiment, the detailed steps of generating the network topological relation graph by adopting the method for discovering the community based on the transformation of the conflict graph are as follows:

(1) obtaining a network graph having v nodes and epsilon links

N-hop conflict graph of

Identification by modularity maximization method

Community C 'of (1)'_m(initialization n ═ 1). Modularity is a measure commonly used to infer the existence of community structures in a network:

wherein e_p＝∑_qE_pqP is the community number, E_ppAs an edge within the community, E_pqTo cross an edge of communities p and q, M represents the expectation that edges located within a community minus edges are randomly distributed without community structure, i.e., M reflects the strength of the community.

(2) Community C 'to Conflict map'_mCarrying out inverse transformation to transform each node into an edge, thereby obtaining the network

Obtaining a plurality of nodes located in a plurality of communities by using a new overlapped community C; the nodes belonging to a plurality of communities are found and only retained in the community set C where the node has the largest number of links.

(3) Using Surrise measure to obtain final community distribution set C obtained after deleting overlapped nodes, and obtaining community distribution C by using direct application module metric maximization algorithm_MAnd comparing and selecting, and selecting the most optimal community of the network partition scheme which maximizes the S value.

(3.1) calculating a Surprise value, and judging the distribution quality of the communities in the network, wherein the distribution quality can be expressed as:

wherein the content of the first and second substances,

the maximum possible number of links in the network, in a k-node network, the value of k (k-1)/2,

For links within communitiesThe maximum value, l is the actual link book in the network, and p is the actual number of links in the community in a partitioning scheme. The S value measures the definite distribution probability of nodes and links in a specific community, and the modularity is maximized when the S ratio is maximized

With better yield, while S depends on the number of links and nodes within each community.

And (3.2) storing the current community set C and the S value thereof in the subnet partition scheme set P.

(3.3) when n is not more than n_maxThen, adjusting n ← n +1, skipping to 1, and regenerating a community distribution set C; when n > n_maxThen, all community graphs and C in the set P are combined_MAnd comparing under different link probabilities, and selecting the most optimal community of the network partitioning scheme which maximizes the S value.

(4) Outputting network topological relation graph with optimal community set C in topological graph form

Network with community set C generated as described above

In the method, the connection relation among all the nodes is only a logical connection relation, and whether the link of wireless communication among the nodes meets the requirement of constructing the communication link is not considered. Therefore, on this basis, the network link is constructed with the objective of optimizing the entire network according to the wireless communication conditions.

On the basis of basic formation of a cluster network structure, in step S3 of this embodiment, a heuristic link construction method combining an average flow capacity ANFC enhancement based on a worldlet feature and cooperative communication is adopted to form a global link connection diagram of a cluster network, so as to generate a cluster initialization network link. Due to the clear tasks, the cluster behavior can be planned, and the space-time distribution and the change can be predicted, so that the cluster network belongs to a deterministic small-world network. To create a deterministic small-world network, certain network characteristics, such as APL, AEL, and network proximity centrality (ACC), can be optimized by adding links. Many deterministic connection addition strategies, such as MinAPL, MaxBC, MaxCC, etc., have a high time complexity, which is typically time critical in drone cluster applications. In the embodiment, by using an average flow capacity enhancement (ACES) heuristic method based on the characteristics of the small world and by using network average flow capacity (ANFC), whether the link establishment brings a capacity improvement effect or not is judged, so that the problems of link transmission congestion and flow load balancing in the link establishment process can be effectively solved, and the link connection diagram of the network bottom layer can be quickly formed.

The Network Flow Capacity (NFC) of the cluster Network as a whole is obtained by summing the maximum Flow Capacity values among all possible nodes in the Network, and for a large-scale cluster (composed of N nodes), the average Network Flow Capacity ANFC is adopted for representation. In a typical link addition strategy based on maximum flow capacity (MaxCap), it often appears that suspended nodes (nodes with only 1 link) in the network are selected for addition to form a minimal cut, which is not meaningful in practical application. According to the embodiment, on the basis of the maximum flow minimum cut theorem, the ANFC is used as a chain establishment judgment standard, and at the moment, the suspended node has little influence on the capacity of the whole network flow, so that an effective link can be established quickly.

In this embodiment, the step of forming a global link connection diagram of a cluster network by using a heuristic link construction method combining the small-world-feature-based average flow capacity ANFC enhancement and cooperative communication to generate a cluster initialization network link includes:

s301, acquiring network initial conditions:

for having N nodes, epsilon links and link capacities of

Initial network of, network

The number of nodes of the link which need to be added is k; first stageInitializing r, node _ set and min-degree, wherein the min _ degree is the degree of the node;

S302, traversing from 1 to N nodes, searching for a node with the degree of p, storing the node in a node _ set, and storing a node with the degree of k and a preset r hop distance in the node _ set (the r value can be set according to application);

s303, after the searching is finished, judging whether a node _ set has a condition meeting the requirement of creating k links, if so, returning to the step S302 to perform next round of traversal searching, and turning to execute the step S304; if not, go to step S305; (ii) a

S304, calculating Euclidean distances among all possible node pairs in the node _ set, arranging the node pairs according to the distance values in a descending order, adding k links for the first k nodes in the node _ set, and adopting a variable capacity C_varAnd a fixed capacity C_fixAfter each link is assigned, adding the direct link graph between the nodes

The above link capacity C (CL)_(i，j)，C_exp) The method mainly comprises the steps that communication signal quality and capacity expectation among nodes are achieved;

s305, if the addition of the direct link construction is not feasible, introducing a cooperative communication relay node for link construction and network supplement communication, and assigning values to link capacity to form a relay graph

In the embodiment, all the suspension nodes and different r-hop neighbors thereof are found out based on ACES to construct link connection in the range, so that the ANFC value can be improved, the r-hop neighbor nodes are introduced according to the characteristics of cluster information transmission requirements, and the constructed network has better applicability. Meanwhile, the complexity of the ACES-based link construction method is O (N) + O (k × N) ²) + O (N), can be simplified to O (k × N)²) Compared with the traditional algorithm such as MaxCap algorithm, the added time complexity is O (k multiplied by N)⁵) The efficiency of the link construction process can be effectively improved, the application requirement of rapid cluster deployment can be met, and the method can also be suitable for the fast deployment of the clusterFixed flow capacity (such as rapid delivery of clusters) and variable flow capacity.

Further, the present embodiment also considers the effect of the Braess paradox (Braess paradox), i.e., if every piece of information in the road network reaches the destination using the most probable path, the time taken to reach the destination may not be optimal, and may even cause a situation that increases network congestion. Therefore, considering the characteristics of schedulable cluster resources, the idle drone can be used as a communication relay, and when the added link does not bring an increase to the ANFC value but reduces the ANFC value, the present embodiment adds a new link by adding a relay node, so as to form a hierarchical network structure of a "layer in which network nodes are connected by directly establishing a link with each other" and a "cooperative communication layer in which a link is established by relaying", as shown in fig. 5, by combining with a distributed routing policy, the situation of overall network congestion can be avoided.

On the basis of a two-layer clustered cluster aerial network structure, considering that the cluster coverage range in practical application is wide, the embodiment further combines a ground high-speed optical fiber network to form a form of immediate access to a ground network in a subarea manner, so that the time information of the cluster and the finger control station is converged and finger control information is issued after landing, and finally a three-layer clustered cluster network structure of an inter-machine network layer, a cooperative communication layer and a ground access layer is formed, as shown in fig. 6.

For a small-world cluster network, group behaviors have certain organizational structures and social characteristics (individual characteristics and social connections among individuals), a large number of local cooperative relationships enable information interaction to be more concentrated in communities or subnetworks of the cluster, and information interaction among the communities or the subnetworks is periodic and sudden. Meanwhile, nodes in the cluster can be scheduled to execute tasks in a cross-region mode, and single-node cross-subnet roaming and multi-node/subnet fusion and analysis phenomena are formed. This makes the design of routing strategy need to take local fast search and instant generation and update of multi-subnet route in node movement into account. The embodiment uses a hierarchical routing strategy facing local aggregation and global roaming, has strong timeliness in the process of transmitting information such as state and cooperation between local nodes, and adopts a forward-looking routing algorithm to search a path locally and quickly; when the remote information transmission such as command control and cluster management is transmitted among roaming nodes which do not belong to the same subnet during the wide-range node roaming across subnets and subnet integration, the delay is tolerable but must be ensured to be reached, and the path is searched in a global orientation mode by adopting a divergent waiting routing algorithm based on Markov prediction based on the movement behavior of the nodes.

In the network generation process, the community generation of the n-hop conflict graph and the establishment of network connection lay a foundation for the initialization of network routing, namely, along with the generation of network topology and links, the adjacent point information of each node and the link relation in the n-hop are initialized in the network generation process, a large number of local connections are established, and the network degree distribution is reasonably controlled, so that complete information is provided for the route initialization generation process. In step 3 of this embodiment, a hierarchical clustering look-ahead routing algorithm is specifically used for communication between nodes, that is, a neighbor node set is constructed based on an n-hop range to obtain a neighborhood, and the target node information is searched for by using the degree of a node as the breadth and using n-hop as the depth. The hierarchical clustering look-ahead routing algorithm specifically comprises the following steps: searching neighbor nodes in n hops, and if a target node is in a source node neighbor domain, directly sending a message to the target node; if not, the source node selects a neighbor as an intermediate forwarding node according to the position of the destination node, and forwards the message to the intermediate forwarding node; and after receiving the message, the intermediate forwarding node judges whether the target node is in the neighbor domain, if so, the intermediate forwarding node sends the message to the target node, otherwise, the intermediate forwarding node continues to select the next intermediate forwarding node and forwards the message until the nodes in the community and the adjacent community are traversed. In the process of searching the adjacency, the relay node of the cooperative communication layer can be used as a neighbor node and listed in a search range, and the relay node continues searching.

In a specific application embodiment, the detailed flow of the look-ahead routing algorithm for hierarchical clustering includes:

(1) inputting: network connection diagram with N nodes

SN Source node, DN destination node, r_lookaheadR hop neighbors of the node, k is the total number of message transmission sessions, Count_sesssionCounting the transmitted message transmission sessions; and (3) initializing: count_sesssion＝1andr_lookahead；

(2) All r-hop neighbors are found and designated as r in the algorithm_lookahead；

(3) If the destination node is in the neighborhood of the source node, the message is directly sent to the destination node; otherwise, switching to (4);

(4) according to the position of the destination node, the source node selects a proper neighbor as an intermediate forwarding node and forwards the message to the forwarding node. When selecting the intermediate forwarding node, an appropriate node is determined according to the degree distribution of the adjacent nodes, the overlapped nodes in the community distribution, the link capacity and the like, namely

(5) After receiving the message, the intermediate forwarding node searches r thereof_lookaheadNeighbor nodes and judging whether a target node is in a neighbor domain, wherein in the process of searching for adjacency, a relay node of the cooperative communication layer can be used as a neighbor node and listed in a search range, and the search is continued through the relay node;

(6) if yes, sending the message to a destination node, otherwise, continuing to select an intermediate node and forwarding, and circularly executing (5) until the destination node is found, and if the destination node is not found, executing (7);

(7) Updating Count_sesssion＝Count_sesssion+1 and routing information table T of all nodes on the routing path_pi。

The time complexity of the distributed routing strategy calculation is approximately O (r × N) in the embodiment²) When the network is constructed, the selection quantity of the next search node is small under the condition of considering node degree distribution and network average degree distribution, and the search scope is effectively limited. Meanwhile, the cluster network is established on the basis of task message community distribution, and the information transmission requirement is concentrated in a community or a subnet/cluster, so that the routing algorithmThe search depth of (a) is also effectively limited.

Analyzing the task execution process and establishing information correlation among the nodes, wherein the node behaviors in the cluster have certain regularity, and the node behaviors which are closely correlated with each other have similarity. On this basis, the embodiment further introduces a markov model, predicts the next hop subnet area of the source node and the destination node, and uses a distribution waiting routing algorithm to inject the message and its copy into the network as much as possible to implement flooding, and the detailed steps are as follows:

firstly, according to the distribution condition of the sub-networks, establishing node accessible Area information Area { C {₁，C₁，...，C_mI.e., area id 1, 2. And (4) counting the regional behaviors accessed by the unmanned aerial vehicle node a, and constructing a region transition probability based on a Markov model. Because each drone records flight information, no additional overhead is incurred to record the area accessed. Let the state set of node a be

(including region ID sequence and number of regions) n +1 st state of node a

Only with the nth state

In this case, the node region transition probability is calculated as follows by using the markov model:

the above probability represents the probability of calculating the transition from state n to n +1 state in the case of transition from state n-1 to n, where X (n-1, n) is the past state transition, where:

the above-mentioned (6) indicates that the number of transitions N (X (N-1, N)) from the state N-1 to the state N accounts for the total number of transitions N (All)_n) The probability of (c).

Then, the next hop regions of the source node and the destination node are predicted to perform routing. When two outbound roaming nodes distributed in different geographic positions or different sub-network regions meet in a certain region, the next hop region of the two meeting nodes is calculated according to the formula by utilizing Markov prediction, if the probabilities calculated by the two nodes in the certain region are both the predicted maximum values, the state transition of the two nodes at the next time is represented, and the nodes in the same region are likely to have the possibility of great overlapping, namely the motion tracks of the two nodes, so that the nodes in the same region only emit one copy, and if the two outbound roaming nodes are not in the same region, L/2 copies are emitted to the meeting nodes. Thus, information and copies are purposefully distributed through location area prediction, so that a rapid search for a transfer path is expected.

In the task execution process, when emergencies such as local severe weather, node battle loss or exit, target threat emergence and the like occur, the unmanned aerial vehicle cluster responds through online decision and task re-planning, so that dynamic changes of a cluster network are caused, network information association and topological relations among nodes are possibly changed, such as unstable or interrupted communication links, changed information flow relations and the like, and the link connection relation, the network topology and even the network hierarchical structure of the original network are required to be adjusted to adapt to the changes of tasks. The stability of the cluster network link determines whether the task can be completed within a rated time and the running safety of the nodes, and for the high-dynamic cluster flight process, the timeliness of link reconstruction also determines the adaptability of the cluster network. In the embodiment, in a hierarchical clustering network structure, link influence entropy (LInE) is used as a criterion of network stability, and a constructed wireless communication model is utilized to settle link quality between nodes in real timeBased on network link maps

Judging the stability of the network; and when the stability is reduced to a preset threshold value within a specified time length, namely suddenly becomes poor, network link reconstruction is triggered, and link reconstruction is carried out by adopting a distributed bee colony algorithm method.

The detailed steps of the link reconfiguration by adopting the distributed swarm algorithm method in the embodiment are as follows:

(1) network stability determination based on LInE behavior in dynamic network

Link influence entropy measures the influence of each link by comparing the change in APL caused by the presence or absence of a link in the network. And calculating the quality of the communication signal according to the position relation between the current nodes if each link exists, and if the basic communication threshold value is not met, determining that the link is disconnected.

At a certain moment, the lene behavior of a link is mainly affected by the network APL and the link location, and can be expressed as:

H＝-∑_i≠jp_ijlog p_ij (8)

wherein H is the network LInE, p_ijIs the influence rate of the link between node pair (i, j), which can be defined as:

wherein the content of the first and second substances,

and in order to remove the APL value of the whole network after the link between the nodes i and i is removed, the denominator is a normalization term, and the APL transformation conditions are summed after the removal. The above formula reflects the actual influence of the link between the nodes i and j, and is calculated by using a wireless communication model based on the relative positions of the nodes and environmental factors.

By acquiring a link influence entropy (LInE) behavior in the cluster dynamic network, the change of the link influence in the network topology can be acquired at different moments, and the stability of the nodes is judged. If the links changing along with the time in the network are less, the nodes are more stable in the network, the cluster topology changes rapidly, the network dynamics is strong, and the key for keeping the network transmission stable is to determine that the influence of the nodes of the stable nodes in the network does not change along with the time and changes frequently.

Stability of influence of a node δ_iDefined as the inverse of the worst case maximum change affected by the node, as follows:

wherein

Representing the influence p of node i_iThe degree to which k deviates from the average at time k, Δ k represents the difference of the dynamic network topology at time k from time k-1,

indicating the maximum average change in node impact at time k.

The abrupt change of the node influence means that the nodes in the network have very strong dynamic property, but the node influence on the stable nodes does not change frequently along with the time change, so the influence stability sigma delta of the nodes involved in the reconstruction process for realizing higher stability of the dynamic network_iShould be minimal.

(2) Network reconstruction based on distributed bee colony heuristic algorithm

The network link reconstruction is carried out by adopting a heuristic algorithm based on the distributed bee colony, and the basic idea is as follows: under the condition that node failure occurs in a cluster network, a potential connectable node is searched by taking the maximum power of the node as a search radius, filtering nodes are screened by taking the connectable time and the connectivity as constraint conditions, then the local optimal solution is updated by continuously updating the neighborhood searching mode of the node to obtain the optimal node, the influence of all node reconstruction links on the network stability is minimized, and the links are connected to complete network reconstruction, so that the global optimal solution is obtained.

The method belongs to the solution of optimization problem, firstly, the mathematical description of the optimization problem is established as follows:

the target is as follows: min sigma_iδ_i，(i∈p_rebuild)

s.t.

Representing a network link map after reconstruction

Representing nodes involved in reconstruction

P_degree≤Max_degreeDenotes the node maximum limit

C_(i，j)≥C_minIndicating that the link capacity setting is greater than the threshold

d_(i，j)≤d_maxAnd means that the distance between the reconstruction nodes is smaller than the maximum communication range.

In the solving process, whether the basic characteristics of the cluster network, such as APL, AEL, ACC and the like, are met or not is considered so as to maintain the characteristics of the cluster network.

In this embodiment, the network link reconfiguration is performed by using a heuristic swarm algorithm, which includes:

s401, when the local communication delay of the network rises to a preset threshold value within a specified time length so as to cause node failure in the network topology, the local communication delay of the network is added to the related node p_iStarting with d_maxSearching the nodes capable of establishing links in the area for the radius, and establishing the population scale m and the maximum iteration number L_maxInitialization traversal number l_sSetting a global variable v, establishing a set of nodes to be reconstructed of the unmanned aerial vehicles with the residual numbers at one time, and guiding all the nodes to complete traversal;

s402, randomly generating initial solutions, recording the adaptive value of each solution, and selecting a neighborhood node with highest fitness as an optimal solution after neighborhood optimization is carried out (by adopting a greedy algorithm);

S403, the local optimization times and a preset threshold value L are set_maxBy comparison, if l_s＞L_maxThen give up whenAnd (3) performing a previous solution, wherein v is v +1, randomly extracting new nodes from the rest bee colony to perform a solution updating process, and juxtaposing traversal times l_s0; if the global variable v is P _ num and P _ num is the number of nodes related to link reconstruction, ending the optimization solution, otherwise, jumping to step S402;

And subsequently, the unmanned aerial vehicle nodes with the numbers of 2-n are respectively used as initial nodes, and the steps are repeatedly executed, so that the link reconstruction of all the nodes can be completed.

In step S403, if the number (degree) of links maintained by the node of the unmanned aerial vehicle is greater than the preset threshold, redundant links between the unmanned aerial vehicles need to be deleted, specifically, δ is deleted when the reachability of the whole node is not changed_iThe smallest link.

The unmanned aerial vehicle cluster is regarded as an intelligent cluster, the task targets are correlated, the task related information interaction is used as interconnection and intercommunication to form a network, the cluster task flow is analyzed from the perspective of social behaviors, a small world network model facing the timely communication of the cluster and multi-dimensional characteristic quantities are established, the community is established according to the information supply and demand relationship, a multi-level and overlapped network structure of the community is established according to the task timing flow, and the cluster behaviors, the task intentions, the information interaction and the network structure generation process can be organically combined.

The foregoing is illustrative of the preferred embodiments of the present invention and is not to be construed as limiting the invention in any way. Although the present invention has been described with reference to the preferred embodiments, it is not intended to be limited thereto. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention shall fall within the protection scope of the technical solution of the present invention, unless the technical essence of the present invention departs from the content of the technical solution of the present invention.

Claims

1. The utility model provides an unmanned aerial vehicle cluster network self-organizing system based on task is cognitive which characterized in that includes:

the physical layer is used for constructing a simulation environment so as to perform simulation test on the performance of the cluster network;

In the application layer, characterizing the cluster tasks based on an information flow dynamic hypergraph model to form an information association graph of each node in the cluster, wherein the information association graph comprises the following steps:

constructing a cluster task space: adopting a hierarchical decomposition type structured modeling method, and according to the sequence from top to bottom, detailing the cluster tasks layer by layer into meta-tasks which are relatively independent and can be directly executed, and mapping the analyzed objects in a hierarchical structure mode;

constructing a dynamic hypergraph model of cluster task information circulation: representing the vertex of the hypergraph by the nodes of the unmanned aerial vehicle, abstracting the meta task into a hyper edge, and connecting the vertex associated with the hyper edge according to the incidence matrix between the meta task and the nodes of the unmanned aerial vehicle; in each hyper-edge, associating the associated vertex by using a connecting line according to the content and the node type of the meta-task to obtain a static hyper-graph model; and finally, arranging the static hypergraph models at a plurality of moments according to a task time sequence to obtain a dynamic time-varying UIF hypergraph model in the whole task process so as to represent the information association relation between nodes and obtain the information association diagram of the nodes.

2. The task awareness-based unmanned aerial vehicle cluster network self-organizing system of claim 1, wherein the application layer comprises:

a topological relation diagram generating module, which is used for establishing information transfer relation between the cluster task information flow and each unmanned aerial vehicle node based on the cluster task represented by the task flow representation module and forming an information association diagram of each node in the cluster

For constructing the network logical topology relationship.

3. The task cognition based unmanned aerial vehicle cluster network self-organizing system of claim 1, wherein the network layer comprises a network topology generating module for constructing the network logical topology relationship according to the information incidence relationship of each node in the cluster; the network topology generation module is a method for discovering communities based on conflict graph transformation, different information services in task information flow are used as requirements of different service communities, node information flow is used as a node logic cross-linking relation in the community, community division is carried out according to logic connection of information transmission, and a community distribution set C is formed_mAnd generating the initial topological graph of the network.

4. The task awareness based drone cluster network self-organizing system of claim 1, wherein the link layer comprises:

the hierarchical routing module is used for constructing a communication routing mode among nodes based on a three-level clustering network structure formed by the hierarchical construction of the link construction module;

5. The method for utilizing the unmanned aerial vehicle cluster network self-organizing system of any one of claims 1-4, wherein the steps comprise:

s1, establishing an information transfer relationship between a cluster task information stream and each unmanned aerial vehicle node according to a cluster task at an application layer to form an information association diagram of each node in a cluster;

s2, at the network layer, constructing a network logic topological relation based on the information association relation of each node in the cluster established by the application layer, and generating a network topological relation graph;

6. The method according to claim 5, wherein in step S1, the cluster tasks are characterized based on an information flow dynamic hypergraph model to form an information association graph of each node in the cluster, and the specific steps include:

s101, cluster task space construction: adopting a hierarchical decomposition type structured modeling method, and according to the sequence from top to bottom, detailing the cluster tasks layer by layer into meta-tasks which are relatively independent and can be directly executed, and mapping the analyzed objects in a hierarchical structure mode;

s102, constructing a cluster task information flow dynamic hypergraph model: representing nodes of the unmanned aerial vehicle as vertexes of a hypergraph, abstracting the meta-task into a hyper-edge, and connecting vertexes associated with the hyper-edge according to an association matrix between the meta-task and the nodes of the unmanned aerial vehicle; in each hyper-edge, associating the associated vertexes by using a connecting line according to the content and the node type of the meta-task to obtain a static hyper-graph model; and finally, arranging the static hypergraph models at a plurality of moments according to a task time sequence to obtain a dynamic time-varying UIF hypergraph model in the whole task process so as to represent the information association relation between nodes and obtain the information association diagram of the nodes.

7. The method according to claim 5, wherein in step S2, a network topology relationship graph is generated by using a method for discovering communities based on conflict graph transformation, wherein after an information association graph between network nodes of a node information flow in a cluster network is obtained, a collision graph mode is used, and a mutual influence relationship between a node and an edge is determined by considering neighborhood influence within n hops of each node; in the process of carrying out community discovery on a conflict graph influenced by n hops, a group of discovered communities comprises different node subsets, wherein each node and at most n-hop neighbors thereof are located in the same subset, and the influence degree information is utilized to identify hierarchical community distribution with larger or smaller scales at different levels;

s201, obtaining a network graph with v nodes and epsilon links

N-hop conflict graph of

Identifying the n-hop conflict graph

Of (5), get the conflict graph community C'_m；

S202, community C 'of conflict graph'_mCarrying out reverse transformation to transform each node into an edge to obtain the network graph

In a new overlapping community, i.e. there is a partial node in A plurality of communities;

s204, the final community distribution set and community distribution C obtained in advance are combined_MComparing, and regenerating a community distribution set according to a comparison result until an optimal community set C is obtained;

and S205, outputting a network topological relation graph with the optimal community set C in a topological graph mode.

8. The method according to claim 5, 6 or 7, wherein in step S3, a heuristic link construction method based on average flow capacity ANFC enhancement combined with cooperative communication is adopted to form a global link connection map of a cluster network to generate cluster initialization network links, and the specific steps include:

s301, acquiring network initial conditions:

is provided with

A node, an epsilon link and a link capacity of

Initial network of, network

The number of nodes of the link which need to be added is k;

s302. from 1 to

Traversing each node, searching for a node with the degree of p, storing the node with the degree of k and the preset r hop distance in a node _ set, and storing the node with the degree of k and the preset r hop distance in the node _ set In the set;

s303, after the searching is finished, judging whether the node _ set has the condition of meeting the k link establishment, if so, returning to the step S302 for the next round of traversal searching, and turning to the step S304; if not, go to step S305;

s304, calculating Euclidean distances among all possible node pairs in the node _ set, arranging the node pairs according to the distance values in a descending order, adding k links for the first k nodes in the node _ set, and adopting variable capacity C_varAnd a fixed capacity C_fixAfter each link is assigned, adding the direct link graph between the nodes

S305, introducing the cooperative communication relay nodes, carrying out link establishment and network supplement communication, assigning values to link capacities, and forming a relay graph

9. The method according to claim 5, 6 or 7, wherein in step S3, the communication between nodes uses a hierarchical clustering look-ahead routing algorithm, that is, a neighbor node set is constructed based on a range of n hops to obtain a neighborhood, and destination node information is searched for using a node degree as a breadth and n hops as a depth, and the specific hierarchical clustering look-ahead routing algorithm specifically includes: searching neighbor nodes in n hops, and if a target node is in a source node neighbor domain, directly sending a message to the target node; if not, the source node selects a neighbor as an intermediate forwarding node according to the position of the destination node, and forwards the message to the intermediate forwarding node; and after receiving the message, the intermediate forwarding node judges whether the target node is in the neighbor domain, if so, the intermediate forwarding node sends the message to the target node, otherwise, the intermediate forwarding node continues to select the next intermediate forwarding node and forwards the message until nodes in the community and the adjacent community are traversed.

10. The method according to claim 5, 6 or 7, wherein in step S4, link quality between nodes is settled in real time by using link influence entropy LInE as a criterion for network stability, and based on network link map

Evaluating the network stability, and triggering network link reconstruction when the stability is reduced to meet a preset condition;

in step S4, performing network link reconfiguration by using a heuristic bee colony algorithm method includes:

s401, when the local communication delay of the network rises to a preset threshold value within a specified time length so as to cause the node failure in the network topology, the related node is subjected to_iThe nodes of the link can be established in the initial search area, and the population scale m and the maximum iteration number L are established_maxInitialization traversal number l_sSetting a global variable v, establishing a set of nodes to be reconstructed of the unmanned aerial vehicles with the residual numbers at one time, and guiding all the nodes to complete traversal;

s403, the local optimization times and a preset threshold value L are set_maxBy comparison, if l_s>L_maxGiving up the current solution, randomly extracting new nodes from the rest bee colony to solve and update, and juxtaposing traversal times l _s0; if the global variable v is P _ num and P _ num is the number of nodes related to link reconstruction, the optimization solution is ended, otherwise, the step S402 is skipped;