Cooperative detection strategy generation method based on soft clustering algorithm

Abstract

According to the soft clustering algorithm-based collaborative detection strategy generation method provided by the invention, task group information can be quickly and effectively generated by adopting a K-means soft clustering algorithm in combination with a collaborative node topological structure and a common link state and a special link state between nodes; according to the method, the aggregation degree between the cooperative nodes and the role importance of the center node and the group main node are fully considered, a decision-making suggestion can be provided for the implementation of multi-node cooperative detection, and the method has the advantages that the clustering degree between the cooperative nodes and the role importance of the center node and the group main node are fully considered; and the method has important guarantee significance for cooperative detection.

Description

Technical Field

[0001] The invention belongs to the technical field of radar task scheduling. Background Art

[0002] As the target detection environment becomes more complex and changeable, single-platform detection can no longer meet actual usage requirements, and multi-node collaboration has become a trend. Collaborative nodes can achieve a variety of tasks such as collaborative search, collaborative tracking, collaborative positioning, collaborative anti-interference, and guided detection. In particular, the reasonable organization and planning of collaborative detection tasks between multiple nodes is of great significance to the improvement of collaborative detection capabilities. In a multi-node detection system, it is generally necessary to specify the role of the node according to the task, and perform different tasks according to different roles. The roles are divided into primary and secondary, and the corresponding task decision-making power and execution power are different.

[0003] The patent "A radar formation collaborative detection method with a multi-task orderly traction strategy" and the patent "Control system and method for collaborative flight of UAV formations based on edge cloud" only perform task allocation and resource scheduling for all nodes in a large formation, without mentioning the division of task groups, and the node roles are only reflected in the master node and the slave node, without a more detailed division. When the collaborative detection system tasks are executed, the task organization and planning are often carried out in a large collaborative network, lacking the division of task groups. The division of appropriate task groups can not only improve the performance of collaborative task execution, but also improve the efficiency of collaborative detection. However, the division criteria of task groups are diverse, and the system robustness is different. In particular, the different choices of central nodes, group master nodes, and task nodes have a serious impact on the implementation of collaborative detection. So, how to reasonably specify the node roles has a great impact on the actual execution effect of collaborative detection tasks. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a collaborative detection strategy generation method based on a soft clustering algorithm, and thereby realizes multi-level collaborative detection between nodes within a group and between task groups. The collaborative detection strategy includes the selection of a task center node, the division of task groups, the selection of a task group master node, and the designation of task nodes. The task center node is the role with the highest priority and decision-making power among all nodes in the collaborative network. It is responsible for completing task planning and task decomposition according to the collaborative detection task application and the collaborative node topology structure, and generating task groups and task group subtasks. The task group master node is the role with the highest priority and decision-making power in a task group. It carries the task node selected by the collaborative detection task group subtask, and generates, evaluates and adjusts the collaborative detection plan of this group according to the collaborative detection subtask. The task node is the specific execution node of the collaborative detection plan. The task node belongs to a certain task group, and there is no task intersection between the nodes of the task groups. The central node and the group master node can also be a task node.

[0005] The present invention proposes a collaborative detection strategy generation method based on a soft clustering algorithm, and the technical solution is as follows:

[0006] Step 1: Collaborative node access to the network;

[0007] Step 2: Central node selection, calculating the central node score selected by each collaborative node in the collaborative network;

[0008] Step 3: Confirm the central node. For a group of central nodes that have been selected, select the node with the largest score as the central node C. cn , where cn is the central node number, and the confirmed central node information is pushed to other nodes in the collaborative network through ordinary links;

[0009] Step 4: Determine the number of task groups K;

[0010] Step 5: Generate K task groups using K-means clustering algorithm;

[0011] Step 6: Elect the group leader node, and calculate the score of the group leader node elected by each collaborative node in the task group;

[0012] Step 7: Group leader node confirmation: For a group of group leader nodes that have been elected, select the node with the largest score as the group leader node G. gn , where gn is the group master node number, and the confirmed group master node information is pushed to other nodes and the central node in the group through ordinary links.

[0013] Step 8: Task node selection: select a node combination that can complete the collaborative detection subtask as the task execution node.

[0014] The beneficial effects of the present invention are:

[0015] The present invention proposes a collaborative detection strategy generation method based on a soft clustering algorithm. The method can combine the collaborative node topology structure and the common link status and special link status between nodes, adopt the K-means soft clustering algorithm to quickly and effectively generate task group information, and adopt a scoring system to complete the central node selection and confirmation as well as the group leader node selection and confirmation, and then the group leader node selects the task execution node. The method fully considers the aggregation degree between collaborative nodes and the role importance of the central node and the group leader node, and can provide a decision-making suggestion for the implementation of multi-node collaborative detection, which has important guarantee significance for collaborative detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a preferred flow chart of the present invention. DETAILED DESCRIPTION

[0017] The present invention is described by way of examples with reference to the accompanying drawings.

[0018] The present invention proposes a collaborative detection strategy generation method based on a soft clustering algorithm, and the preferred implementation process is as follows: Figure 1 , specifically described as the following process:

[0019] Step 1: The collaborative node enters the network. Let the number of collaborative nodes in the network be N, and the link state between the collaborative nodes be nl. i,j (s c ,w c ,s s ,w s ) including ordinary links c and special links s , and share each other's location information through common links gl(n,lon,lat,h), where n is the collaborative node number, lon is the node longitude, lat is the node latitude, h is the node height, s c is the effective state of the common link between the i-th node and the j-th node, s c =1 means valid, s c =0 means invalid, w c is the transmission bandwidth between two nodes when the normal link is valid; s s is the effective state of the special link between the i-th node and the j-th node, s s =1 means valid, s s =0 means invalid, w s is the transmission bandwidth between two nodes when a special link is valid;

[0020] Step 2: Central node selection, calculating the central node score selected by each collaborative node in the collaborative network;

[0021] The optimization process is as follows: let the node number be k n , 1≤n≤N, the kth n The central node numbered by the nodes is i n , calculate the i-th n The sum of the distances between a node and other nodes in the collaborative network Get the i n The distance state factor of each node Where r(i n ,j) is the n The longitude and latitude of the node gl(i n ,lon,lat,h) and the j-th node gl(j,lon,lat,h) to calculate the distance between the two nodes; calculate the i-th node n The sum of the valid status values ​​of the common links between the node and other nodes Get the i n The special link state factor between a node and other nodes Calculate the i n The state score of a node as a central node is C c (i n )=λ1R e (i n )+λ2S e (i n ), where λ1, λ2 are weighting coefficients, and λ1+λ2=1;

[0022] Step 3: Confirm the central node. For a group of central nodes that have been selected, select the node with the largest score as the central node C. cn , where cn is the central node number, and the confirmed central node information is pushed to other nodes in the collaborative network through ordinary links;

[0023] Step 4: Determine the number of task groups K, and set K to be in the range of 1 to N. The preferred method is as follows:

[0024] Step 4-1, for each K value, execute the K-means clustering algorithm;

[0025] Step 4-1-1, initialize K cluster centers;

[0026] Step 4-1-2, assign the central node to each group;

[0027] Step 4-1-3, assign nodes to the nearest cluster center;

[0028] Step 4-1-4, update the cluster center;

[0029] Step 4-1-5: Repeat the above steps until convergence.

[0030] Step 4-2, for each K value, calculate the sum of the squared distances;

[0031] Step 4-2-1, calculate the sum of the squares of the distances from each node in each group to the center of the cluster to which it belongs;

[0032] Step 4-2-2: Add the sum of the squared distances of each group to get the sum of the squared distances ΓR 2 (k), k∈[1,N].

[0033] Step 4-3: For a set of ΓR 2 (k), find the optimal K value;

[0034] Step 4-3-1, for k∈[2,N-1], calculate the slope change rate of its two adjacent points respectively;

[0035] Step 4-3-2: Select the k value corresponding to the point with the largest slope change rate as the optimal K value.

[0036] Step 5: Use the K-means clustering algorithm to generate K task groups. The specific method is as follows:

[0037] Step 5-1, initialize K cluster centers;

[0038] Step 5-2, assign the central node to each group;

[0039] Step 5-3, assign nodes to the nearest cluster center;

[0040] Step 5-4, update the cluster center;

[0041] Step 5-5: Repeat the above steps until convergence.

[0042] Step 6: Elect the group leader node, and calculate the score of the group leader node elected by each collaborative node in the task group;

[0043] The preferred method is as follows: The central node is known as C cn , the number of nodes in the kth, 1≤k≤K task group is M, and the node number in the group is k m , 1≤m≤M, let the kth m The group leader node numbered by the nodes is i m , calculate the i-th m The sum of the distances between a node and other nodes in the task group Get the i m The distance state factor of each node Where r(i m ,j) is the m The longitude and latitude of the node gl(i m ,lon,lat,h) and the j-th node gl(j,lon,lat,h) to calculate the distance between the two nodes; calculate the i-th node m The sum of the effective state values ​​of the special links between the node and other nodes Get the i m The special link state factor between a node and other nodes Calculate the i m The status score of the node as the leader node where s s (i m ,cn) is the i-th m The special link status of each node and the central node, η1, η2 are weighted coefficients, and η1+η2=1.

[0044] Step 7: Group leader node confirmation: For a group of group leader nodes that have been elected, select the node with the largest score as the group leader node G. gn, where gn is the group master node number, and the confirmed group master node information is pushed to other nodes and the central node in the group through ordinary links.

[0045] Step 8: Task node selection: The group leader node selects a node combination that can complete the collaborative detection subtask as the task execution node based on the collaborative detection plan and the link status and topological relationship between the nodes in the group. The judgment method for completing the collaborative detection subtask includes but is not limited to whether the node detection power covers the detection target and whether the node detection data rate meets the task requirements.

[0046] The present invention fully considers the aggregation degree between collaborative nodes and the role importance of the central node and the group leader node, and can provide a decision-making suggestion for the implementation of multi-node collaborative detection, which has an important guarantee significance for collaborative detection.

Claims

1. A collaborative detection strategy generation method based on a soft clustering algorithm, characterized in that: Step 1: The collaborative node enters the network. Let the number of collaborative nodes in the network be N, and the link state between the collaborative nodes be nl. i,j (s c ,w c ,s s ,w s ) including ordinary links c and special links s , and share each other's location information through common links gl(n,lon,lat,h), where n is the collaborative node number, lon is the node longitude, lat is the node latitude, h is the node height, s c is the effective state of the common link between the i-th node and the j-th node, s c =1 means valid, s c =0 means invalid, w c is the transmission bandwidth between two nodes when the normal link is valid; s s is the effective state of the special link between the i-th node and the j-th node, s s =1 means valid, s s =0 means invalid, w s is the transmission bandwidth between two nodes when a special link is valid; Step 2: Central node selection, calculating the central node score selected by each collaborative node in the collaborative network; Step 3: Confirm the central node. For a group of central nodes that have been selected, select the node with the largest score as the central node C. cn , where cn is the central node number, and the confirmed central node information is pushed to other nodes in the collaborative network through ordinary links; Step 4: Determine the number of task groups K; Step 5: Generate K task groups using K-means clustering algorithm; Step 6: Elect the group leader node, and calculate the score of the group leader node elected by each collaborative node in the task group; Step 7: Group leader node confirmation: For a group of group leader nodes that have been elected, select the node with the largest score as the group leader node G. gn , where gn is the group master node number, and the confirmed group master node information is pushed to other nodes and the central node in the group through the common link; Step 8: Task node selection: select a node combination that can complete the collaborative detection subtask as the task execution node.

2. The method for generating a collaborative detection strategy based on a soft clustering algorithm according to claim 1, characterized in that: The step 2 also includes: let the node number be k n , 1≤n≤N, the kth n The central node numbered by the nodes is i n , calculate the i-th n The sum of the distances between a node and other nodes in the collaborative network Where r(i n ,j) is the i-th n The distance between the i-th node and the j-th node is obtained n The distance state factor of each node Calculate the i n The sum of the valid status values ​​of the common links between the node and other nodes Get the i n The common link state factor between a node and other nodes Calculate the i n The state score of a node as a central node is C c (i n )=λ1R e (i n )+λ2S e (i n ), where λ1, λ2 are weighting coefficients, and λ1+λ2=1.

3. The method for generating a collaborative detection strategy based on a soft clustering algorithm according to claim 1, characterized in that: In step 4, the number of task groups K is determined, and K is set in the range of 1 to N, including: Step 4-1, for each K value, execute the K-means clustering algorithm; Step 4-1-1, initialize K cluster centers; Step 4-1-2, assign the central node to each group; Step 4-1-3, assign nodes to the nearest cluster center; Step 4-1-4, update the cluster center; Step 4-1-5, repeat the above steps until convergence; Step 4-2, for each K value, calculate the sum of the squared distances; Step 4-2-1, calculate the sum of the squares of the distances from each node in each group to the center of the cluster to which it belongs; Step 4-2-2: Add the sum of the squared distances of each group to get the sum of the squared distances ΓR 2 (k), k∈[1,N]; Step 4-3: For a set of ΓR 2 (k), find the optimal K value; Step 4-3-1, for k∈[2,N-1], calculate the slope change rate of its two adjacent points respectively; Step 4-3-2: Select the k value corresponding to the point with the largest slope change rate as the optimal K value.

4. The method for generating a collaborative detection strategy based on a soft clustering algorithm according to claim 1, characterized in that: The step 6 also includes: knowing that the central node is C cn , the number of nodes in the kth, 1≤k≤K task group is M, and the node number in the group is k m , 1≤m≤M, let the kth m The group leader node numbered by the nodes is i m , calculate the i-th m The sum of the distances between a node and other nodes in the task group Where r(i m ,j) is the i-th m The distance between the i-th node and the j-th node is obtained m The distance state factor of each node Calculate the i m The sum of the effective state values ​​of the special links between the node and other nodes Get the i m The special link state factor between a node and other nodes Calculate the i m The status score of the node as the leader node where s s (i m ,cn) is the i-th m The special link status of each node and the central node, η1, η2 are weighted coefficients, and η1+η2=1.

Images