Three-dimensional wireless sensor network node deployment method

Abstract

The invention discloses a three-dimensional wireless sensor network node deployment method, which comprises the following steps of: firstly, selecting a measurement region for initialization, and dividing the space in which the measurement region is located into sub-regions; secondly, performing node initialization deployment on the sub-region space, improving the initialization process by adopting a greedy strategy, quickly calculating an uncovered region of the sub-region by utilizing a Monte Carlo method, calculating the distance between a cluster center node and other nodes in the sub-region, and optimizing the initial position of the node according to a tabu search target function; and finally, optimizing the whole node according to the Euclidean distance of the cluster center node and the occupancy ratio of the cluster center node in the sub-region so as to adjust the working state of each node in the whole region and the sub-region. According to the method, the uncertainty generated by the randomization of the initial deployment of the nodes is comprehensively considered, and the improved K-means algorithm is utilized to quickly provide the clustering center, so that the coverage rate of the nodes of the three-dimensional wireless sensor network can be effectively improved, and the problem of waste generated by excessive deployment of the nodes can be reduced.

Description

Technical field:

[0001] The present invention belongs to the technical field of wireless sensor networks, and in particular relates to a three-dimensional wireless sensor network node deployment method based on an improved K-means algorithm with a greedy strategy, which is used to improve network coverage and energy efficiency. Background technology:

[0002] As the complexity of environmental monitoring tasks continues to increase, people have higher requirements for the comprehensiveness and accuracy of data information. In order to meet these requirements, people usually use sensors with perception capabilities as nodes to build a spatial network to perform data collection and transmission tasks.

[0003] Given the random deployment and widespread distribution of nodes in wireless sensor networks, the communication distance and signal strength between nodes are easily affected by many factors in practical applications. In order to ensure the reliability and stability of the network and effectively improve network performance, it is of great significance to study an appropriate coverage control method to improve the monitoring effect of directed sensor networks. To this end, researchers have proposed a variety of optimization algorithms to improve the efficiency and accuracy of data transmission.

[0004] So far, research on wireless sensor network node deployment methods has mainly focused on two-dimensional environments. In the Acta Physica Sinica, a paper written by scholar Fang Wei published "Wireless Sensor Network Coverage Control Deployment Strategy Based on Voronoi Diagram Blind Areas". This strategy first divides the monitoring area into Voronoi diagrams to determine the Thiessen polygons covered by each sensor node. Subsequently, by analyzing the coverage of the vertices, the blind areas within the Thiessen polygons are identified, and their geometric centers are constructed as potential target locations for sensor node movement, aiming to improve the coverage of the network. However, the research on this method is limited to two-dimensional environments, and the node perception model has limitations. In addition, over-emphasizing overall node deployment may lead to a decrease in the measurement accuracy of local node deployment, which in turn affects the coverage of the sensor network in the measurement area.

[0005] Scholars Liu Cuiping et al. proposed a sensor node deployment strategy based on the firefly swarm optimization algorithm in the paper "Wireless Sensor Node Deployment Based on Firefly Swarm Optimization Algorithm". This strategy first randomly deploys sensor nodes, and then determines the node's movement direction based on the calculated movement probability, so as to achieve effective node deployment. However, this method still faces the problem of excessive number of node deployment and coverage blind spots in practical applications.

[0006] In the paper "Research on Patent Map Production Method Based on K-means Clustering Algorithm", scholar Qiu Honghua and his colleagues proposed a patent map production structure optimization method based on K-means clustering algorithm. The method first classifies the entries of the patent documents involved in the study, including structured and unstructured items, and constructs a semantic network through the K-means clustering algorithm to complete the production of the patent map. However, the process of constructing a keyword matrix and performing clustering classification in the initial stage of this method is relatively complicated, resulting in the inability to quickly determine the cluster center, increasing the difficulty of achieving the goal.

[0007] Therefore, a three-dimensional wireless sensor network node deployment method is needed that can comprehensively consider the communication efficiency and coverage between local area sensor network nodes and the overall area.

[0008] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention:

[0009] The purpose of the present invention is to provide a three-dimensional wireless sensor network node deployment method, which divides the measurement area, adopts a greedy strategy to improve the initialization process, uses a Monte Carlo method to quickly calculate the uncovered area of ​​the sub-area, calculates the distance between the cluster center node and other nodes in the sub-area, and optimizes the node initial position according to the taboo search objective function; uses a K-means algorithm to associate nodes in different sub-areas to determine the optimal solution between and within the area; in addition, uses a hierarchical splitting concept to divide the data set into several smaller clusters, and uses the K-means algorithm to calculate the evaluation function value to decide whether to merge the clusters, and then iterates and updates until the number of iterations or the set threshold is met, thereby optimizing the three-dimensional wireless sensor network node deployment, thereby overcoming the defects in the above-mentioned prior art.

[0010] To achieve the above object, the present invention provides a three-dimensional wireless sensor network node deployment method, the steps of which are:

[0011] S01. In the three-dimensional measurement area, a rectangular measurement area is selected to establish a three-dimensional rectangular coordinate system, and the measurement area is divided into three-dimensional grids, and randomly divided into M sub-areas to obtain a sub-area set A. i , where i = 1, 2, 3, ..., M;

[0012] S02. Initialize node deployment in each sub-area, and use a greedy strategy to deploy nodes in each sub-area so that each sub-area can be fully covered by the nodes in the area, and obtain the sub-area node set S i, and optimize the node position according to the taboo search objective function, where is a node in sub-region i, j is the number of nodes in the sub-region; node Including communication radius and the three-dimensional coordinates of its own position are the vertical distances from the node to the x-axis, y-axis, and z-axis of the coordinate system respectively;

[0013] S03, dividing the sub-region sets into different clusters, determining the cluster center nodes in different clusters by using the K-means algorithm, and setting a threshold ε of the distance between the remaining nodes and the cluster center node; judging the threshold ε of the distance between the remaining nodes and the cluster center node and iteratively optimizing according to the result, adjusting the node position, so that the position of the selected cluster center node is closer to the minimum average position of the nodes in the cluster, and reducing the overall distance deviation;

[0014] S04. Calculate the cluster center node occupancy ratio η in each sub-region N ,The cluster center node occupancy ratio is used to determine whether to adjust the node position;

[0015]

[0016] Among them, ω N is the cluster center node S N The number of all centroids perceived, ξ N is the number of nodes that are not sensed by the cluster center node.

[0017] Preferably, in the technical solution, the greedy strategy process of step S02 is:

[0018] S21. For each sub-region, the Monte Carlo method is used to estimate the uncovered locations of the nodes in the current region, and a large number of random point sets are generated in the sub-region;

[0019] The distance from a random point to a node can be calculated using the Euclidean distance formula:

[0020]

[0021] Where dist is the distance from the random point to the deployed node. If dist is less than or equal to the communication radius If , the random point is inside the ball, otherwise, the random point is not inside the ball;

[0022] S22. Count the number of random points in the ball and calculate the proportion η of the total number of random points in the sub-area. count , and multiply the ratio by the volume of the sub-region cuboid to obtain the region coverage η cover ;

[0023] ηcover =η count ×px area × area ×pz area (3)

[0024] Among them, px area is the value of the three-dimensional measurement area on the x-axis of the coordinate system, py area is the value of the three-dimensional measurement area on the y-axis of the coordinate system, pz area is the value of the three-dimensional measurement area on the z-axis of the coordinate system;

[0025] S23, selecting an uncovered position in the sub-area, and placing a new node at the position so that the coverage of the new node is maximized;

[0026] S24, sub-region node set S i As the initial solution of the tabu search, when the node When moving a γ, if the node satisfies the objective function λ greater than 1 / 2, the node position is updated to the coordinates after the move, otherwise, it is unchanged, the taboo table records the node, and traverses the remaining nodes;

[0027]

[0028] Among them, η new is the coverage rate of the new covered sub-region space, η other is the coverage ratio of the space covered by other nodes;

[0029]

[0030] Among them, γ is defined as the node A small step in position movement;

[0031] S25. Repeat steps S21-S24 to perform iterative update, update the sub-regions, until all sub-regions are completely covered by the nodes of the region and the initial position is optimized.

[0032] Preferably, in the technical solution, the iterative process of the K-means algorithm in step S03 is:

[0033] S31, the sub-region selects random nodes in the region as the initial cluster center nodes, and obtains the initial cluster center node set C = {S1, S2, ..., S N}, where S N represents the initial cluster center node of the sub-region, and N represents the number of initial cluster center nodes of the sub-region;

[0034] S32, for each node, calculate its distance from the central node of each initial cluster, and assign it to the nearest cluster;

[0035] Among them, the average Euclidean distance formula between the node and each cluster center node is:

[0036]

[0037] Among them, D i is the average Euclidean distance between the node and each cluster center node, dists is the distance function from the rest of the nodes to the cluster center node after randomization, and Count is the sum of the initial number of cluster center nodes in the randomly generated sub-regions in the region;

[0038] dists=dist i ×c1 (7),

[0039] Among them, dist i is the Euclidean distance from the node to the center nodes of other clusters, and the random factor c1 is a number between [0.9, 1] generated by the function random;

[0040] S33, if the difference between the distance between the cluster center node of the cluster and two nodes other than the cluster center node is less than the set threshold ε, the two nodes are deleted, and a new node is added so that the distance to the cluster center node is equal and the sum of the distances to the original positions of the two nodes is the shortest;

[0041] S34. Repeat the above steps S32-S33 until the maximum number of iterations is reached or the execution positions of all nodes do not change.

[0042] Preferably, in the technical solution, the process of determining the cluster center node occupancy ratio in step S04 is as follows:

[0043] S41, for each cluster, update the cluster center node of each cluster; let all nodes in each cluster be the cluster center nodes in their clusters again, and calculate their cluster center node occupancy ratios in turn;

[0044] S42, calculating the cluster center node occupancy ratio of all nodes in each sub-region, and taking the cluster center node with the lowest occupancy ratio in the region as the new cluster center node;

[0045] S43, repeating the above steps S41-S42 until the positions of all cluster center nodes no longer change or the maximum number of iterations is reached, and obtaining the final cluster center node position set dots.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention divides the area to be tested into various small intervals, deploys node initialization according to the greedy strategy, uses the Monte Carlo method to quickly estimate the space utilization of the initialized nodes and reduces the significant impact of the initialization randomness on the subsequent judgment iteration, so that each sub-area is completely covered. According to the sub-area set, it is divided into different clusters, the center point of each cluster is determined, and the problem of falling into the local optimal solution cycle is reduced by introducing a random factor, and the threshold of the distance between the node and the center point is set. The result is iteratively optimized by the K-means algorithm to adjust the node position. The position of sensor nodes is adjusted by using the cluster center node occupancy ratio to optimize the node deployment between sub-areas. The problem that the K-means algorithm randomly gives the initial cluster center when initializing clustering, which causes the uncertainty caused by the randomization of node initialization deployment to cause large fluctuations in the clustering results, making the clustering results unstable and prone to falling into local optimality and thus ignoring the overall situation, is overcome. The K-means algorithm improved by the greedy strategy has low computational time complexity and fast convergence speed, which can reduce the significant impact of initialization randomness on subsequent judgment iterations, effectively improve deployment efficiency and wireless sensor network coverage, and optimize the traditional K-means algorithm. The problem that the randomization of node initialization deployment produces great uncertainty and has a great impact on the selection of initial cluster centers on the clustering results ensures that the reliability and stability of the network are improved. Description of the drawings:

[0048] Figure 1 A flow chart of a three-dimensional wireless sensor network node deployment method of the present invention;

[0049] Figure 2 It is a schematic diagram of the overall regional division of the present invention;

[0050] Figure 3 This is a schematic diagram of node deployment in a single area of ​​the present invention;

[0051] Figure 4 Schematic diagram of node determination division between different regions of the present invention. Specific implementation method:

[0052] The specific embodiments of the present invention are described in detail below, but it should be understood that the protection scope of the present invention is not limited by the specific embodiments.

[0053] Unless explicitly stated otherwise, throughout the specification and claims, the term “comprise” or variations such as “include” or “comprising”, etc., will be understood to include the stated elements or components but not to exclude other elements or components.

[0054] like Figure 1 As shown, a three-dimensional wireless sensor network node deployment method, the steps are:

[0055] S01. In the three-dimensional measurement area, a rectangular measurement area is selected to establish a three-dimensional rectangular coordinate system, and the measurement area is divided into three-dimensional grids, and randomly divided into M sub-areas to obtain a sub-area set A. i , where i = 1, 2, 3, ..., M;

[0056] For example, Figure 2 As shown, in the three-dimensional measurement area, a cuboid is selected as the measurement area, and the space therein is divided, and the number of small cuboids in the sub-area is 320, the number of nodes is 100, and the node communication radius is selected as 50;

[0057] S02. Initialize node deployment in each sub-area, and use a greedy strategy to deploy nodes in each sub-area so that each sub-area can be fully covered by the nodes in the area, and obtain the sub-area node set S i , and optimize the node position according to the taboo search objective function, where is a node in sub-region i, j is the number of nodes in the sub-region; node Including communication radius and the three-dimensional coordinates of its own position are the vertical distances from the node to the x-axis, y-axis, and z-axis of the coordinate system respectively;

[0058] S21. For each sub-region, the Monte Carlo method is used to estimate the uncovered locations of the nodes in the current region, and a large number of random point sets are generated in the sub-region;

[0059] The distance from a random point to a node can be calculated using the Euclidean distance formula:

[0060]

[0061] Where dist is the distance from the random point to the deployed node. If dist is less than or equal to the communication radius If , the random point is inside the ball, otherwise, the random point is not inside the ball;

[0062] S22. Count the number of random points in the ball and calculate the proportion η of the total number of random points in the sub-area. count and used in communication radius The ratio of random points in the ball multiplied by the volume of the sub-region cuboid gives the region coverage η cover ;

[0063] η cover =η count ×px area × area ×pz area (3)

[0064] Among them, px area is the value of the three-dimensional measurement area on the x-axis of the coordinate system, py area is the value of the three-dimensional measurement area on the y-axis of the coordinate system, pz area is the value of the three-dimensional measurement area on the z-axis of the coordinate system;

[0065] S23, selecting an uncovered position in the sub-area, and placing a new node at the position so that the coverage of the new node is maximized;

[0066] like Figure 3 As shown in FIG. 1 , the number of random points in the ball in sub-region i is 2100, and the total number of random points generated in the sub-region is 21000. Then η cover It is 2100 / 21000, and the coverage is 0.1. It is necessary to repeat the steps to continue generating new nodes;

[0067] S24, sub-region node set S i As the initial solution of the tabu search, when the node When moving a γ, if the node satisfies the objective function λ greater than 1 / 2, the node position is updated to the coordinates after the move, otherwise, it is unchanged, the taboo table records the node, and traverses the remaining nodes;

[0068]

[0069] Among them, η new is the coverage rate of the new covered sub-region space, η other is the coverage ratio of the space covered by other nodes;

[0070]

[0071] Among them, γ is defined as the node A small step in position movement;

[0072] S25, repeating steps S21-S24 to perform iterative update, updating the sub-regions until all sub-regions are completely covered by the nodes of the region and the initial positions are optimized;

[0073] S03, dividing the sub-region sets into different clusters, determining the cluster center nodes in different clusters by using the K-means algorithm, and setting a threshold ε of the distance between the remaining nodes and the cluster center node; judging the threshold ε of the distance between the remaining nodes and the cluster center node and iteratively optimizing according to the result, adjusting the node position, so that the position of the selected cluster center node is closer to the minimum average position of the nodes in the cluster, and reducing the overall distance deviation;

[0074] S31, the sub-region selects random nodes in the region as the initial cluster center nodes, and obtains the initial cluster center node set C = {S1, S2, ..., S N}, where S N represents the initial cluster center node of the sub-region, and N represents the number of initial cluster center nodes of the sub-region;

[0075] S32, for each node, calculate its distance from the central node of each initial cluster, and assign it to the nearest cluster;

[0076] Among them, the average Euclidean distance formula between the node and each cluster center node is:

[0077]

[0078] Among them, D i is the average Euclidean distance between the node and each cluster center node, dists is the distance function from the rest of the nodes to the cluster center node after randomization, and Count is the sum of the initial number of cluster center nodes in the randomly generated sub-regions in the region;

[0079] dists=dist i ×c1 (7),

[0080] Among them, dist i is the Euclidean distance from the node to the center nodes of other clusters, and the random factor c1 is a number between [0.9, 1] generated by the function random;

[0081] S33, if the difference between the distance between the cluster center node of the cluster and two nodes other than the cluster center node is less than the set threshold ε, the two nodes are deleted, and a new node is added so that the distance to the cluster center node is equal and the sum of the distances to the original positions of the two nodes is the shortest;

[0082] like Figure 4 As shown, the location of the cluster center node is known (3, 6, 7), and the threshold is set to 1 unit length. There are two node locations (6, 11, 11) and (8, 9, 11). It can be concluded that the difference in distance between the two other points is 0, which is less than the set threshold, and a new node is generated for replacement.

[0083] S34, repeat the above steps S32-S33 until the maximum number of iterations is reached or the execution positions of all nodes do not change;

[0084] S04. Calculate the cluster center node occupancy ratio η in each sub-region N ,The cluster center node occupancy ratio is used to determine whether to adjust the node position;

[0085]

[0086] Among them, ω N is the cluster center node S N The number of all centroids perceived, ξ N is the number of nodes that are not sensed by the cluster center node.

[0087] S41, for each cluster, update the cluster center node of each cluster; let all nodes in each cluster be the cluster center nodes in their clusters again, and calculate their cluster center node occupancy ratios in turn;

[0088] S42, calculating the cluster center node occupancy ratio of all nodes in each sub-region, and taking the cluster center node with the lowest occupancy ratio in the region as the new cluster center node;

[0089] S43, repeating the above steps S41-S42 until the positions of all cluster center nodes no longer change or the maximum number of iterations is reached, and obtaining the final cluster center node position set dots.

[0090] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A three-dimensional wireless sensor network node deployment method, the steps of which are: S01. In the three-dimensional measurement area, a rectangular measurement area is selected to establish a three-dimensional rectangular coordinate system, and the measurement area is divided into three-dimensional grids, and randomly divided into M sub-areas to obtain a sub-area set A. i , where i = 1, 2, 3, ..., M; S02. Initialize node deployment in each sub-area, and use a greedy strategy to deploy nodes in each sub-area so that each sub-area can be fully covered by the nodes in the area, and obtain the sub-area node set S i , and optimize the node position according to the taboo search objective function, where is a node in sub-region i, j is the number of nodes in the sub-region; node Including communication radius and the three-dimensional coordinates of its own position are the vertical distances from the node to the x-axis, y-axis, and z-axis of the coordinate system respectively; S03, dividing the sub-region sets into different clusters, determining the cluster center nodes in different clusters by using the K-means algorithm, and setting a threshold ε of the distance between the remaining nodes and the cluster center node; judging the threshold ε of the distance between the remaining nodes and the cluster center node and iteratively optimizing according to the result, adjusting the node position, so that the position of the selected cluster center node is closer to the minimum average position of the nodes in the cluster, and reducing the overall distance deviation; S04. Calculate the cluster center node occupancy ratio η in each sub-region N ,The cluster center node occupancy ratio is used to determine whether to adjust the node position; Among them, ω N is the cluster center node S N The number of all centroids perceived, ξ N is the number of nodes that are not sensed by the cluster center node.

2. A three-dimensional wireless sensor network node deployment method according to claim 1, characterized in that: The greedy strategy process of step S02 is: S21. For each sub-region, the Monte Carlo method is used to estimate the uncovered locations of the nodes in the current region, and a large number of random point sets are generated in the sub-region; The distance from a random point to a node can be calculated using the Euclidean distance formula: Where dist is the distance from the random point to the deployed node. If dist is less than or equal to the communication radius If , the random point is inside the ball, otherwise, the random point is not inside the ball; S22. Count the number of random points in the ball and calculate the proportion η of the total number of random points in the sub-area. count , and multiply the ratio by the volume of the sub-region cuboid to obtain the region coverage η cover ; or cover =the count ×px area ×py area ×pz area (3), Among them, px area is the value of the three-dimensional measurement area on the x-axis of the coordinate system, py area is the value of the three-dimensional measurement area on the y-axis of the coordinate system, pz area is the value of the three-dimensional measurement area on the z-axis of the coordinate system; S23, selecting an uncovered position in the sub-area, and placing a new node at the position so that the coverage of the new node is maximized; S24, sub-region node set S i As the initial solution of the tabu search, when the node When moving a γ, if the node satisfies the objective function λ greater than 1 / 2, the node position is updated to the coordinates after the move, otherwise, it is unchanged, the taboo table records the node, and traverses the remaining nodes; Among them, η new is the coverage rate of the new covered sub-region space, η other is the coverage ratio of the space covered by other nodes; Among them, γ is defined as the node A small step in position movement; S25. Repeat steps S21-S24 to perform iterative update, update the sub-regions, until all sub-regions are completely covered by the nodes of the region and the initial position is optimized.

3. A three-dimensional wireless sensor network node deployment method according to claim 1, characterized in that: The iterative process of the K-means algorithm in step S03 is: S31, the sub-region selects random nodes in the region as the initial cluster center nodes, and obtains the initial cluster center node set C = {S1, S2, ..., S N }, where S N represents the initial cluster center node of the sub-region, and N represents the number of initial cluster center nodes of the sub-region; S32, for each node, calculate its distance from the central node of each initial cluster, and assign it to the nearest cluster; Among them, the average Euclidean distance formula between the node and each cluster center node is: Among them, D i is the average Euclidean distance between the node and each cluster center node, dists is the distance function from the rest of the nodes to the cluster center node after randomization, and Count is the sum of the initial number of cluster center nodes in the randomly generated sub-regions in the region; dists=dist i ×c1(7), Among them, dist i is the Euclidean distance from the node to the center nodes of other clusters, and the random factor c1 is a number between [0.9, 1] generated by the function random; S33, if the difference between the distance between the cluster center node of the cluster and two nodes other than the cluster center node is less than the set threshold ε, the two nodes are deleted, and a new node is added so that the distance to the cluster center node is equal and the sum of the distances to the original positions of the two nodes is the shortest; S34. Repeat the above steps S32-S33 until the maximum number of iterations is reached or the execution positions of all nodes do not change.

4. A three-dimensional wireless sensor network node deployment method according to claim 1, characterized in that: The process of determining the cluster center node occupancy ratio in step S04 is as follows: S41, for each cluster, update the cluster center node of each cluster; let all nodes in each cluster be the cluster center nodes in their clusters again, and calculate their cluster center node occupancy ratios in turn; S42, calculating the cluster center node occupancy ratio of all nodes in each sub-region, and taking the cluster center node with the lowest occupancy ratio in the region as the new cluster center node; S43, repeating the above steps S41-S42 until the positions of all cluster center nodes no longer change or the maximum number of iterations is reached, and obtaining the final cluster center node position set dots.

Images