Optimized clustering routing design method applied to WSN (Wireless Sensor Network)

Abstract

The invention belongs to the technical field of Internet of Things communication, and particularly discloses an optimized clustering routing design method applied to a WSN (Wireless Sensor Network), which adopts the specific technical scheme that an improved spectral clustering method is utilized to perform clustering operation on a WSN node structure, a transmission path of WSN nodes is preliminarily planned, a shortest path from each point to a cluster head is constructed, and a clustering result is obtained; according to the two improved methods, factors such as node residual energy, distance, load, relay node selection times and the like are comprehensively considered, a final value is solved by combining a weight value of a CRITIC design edge to replace an own edge weight value of a Dijkstra and an undirected graph in spectral clustering, and finally, respective energy consumption values of each Dijkstra algorithm transmission path and a point-to-point transmission path are compared to obtain a final value. According to the method, the energy consumption during data transmission is reduced, the life cycle of the nodes is prolonged, the death rate is reduced, and the energy balance of the WSN is enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet of Things communications, and in particular relates to an optimized cluster routing design method applied in WSN. Background Art

[0002] In most wireless sensor networks, due to the influence of price and hardware size, the communication processing ability and computing ability of sensor nodes are poorer and the storage capacity is smaller. Sensor nodes are generally deployed in large numbers in remote and severe environments by means of aircraft scattering or manual arrangement. Especially in the case of large-scale WSN, there will be the following problems:

[0003] First, batteries are often used for power supply, which has very limited energy. If data transmission is random or has inappropriate rules, the system life cycle will be very short.

[0004] Second, in traditional WSNs, due to the uneven distribution of communication and computation tasks, some nodes may run out of energy faster than others, resulting in uneven network coverage and potential communication holes.

[0005] Third, without an effective routing strategy, packets may undergo unnecessary multiple forwardings, which not only increases latency but also exacerbates energy consumption.

[0006] Fourth, nodes in WSN may change their locations due to environmental changes or mobility, and traditional methods may have difficulty in quickly adapting to such dynamics. Summary of the invention

[0007] In order to solve the technical problems existing in the prior art, the present invention proposes an optimized clustering routing design method applied to WSN. Spectral clustering is a clustering algorithm based on graph theory. It uses the Laplace matrix of the network to perform eigenvalue decomposition to discover the low-dimensional structure of data. In WSN, spectral clustering can not only be used for node classification, but also for network management. Dijkstra algorithm is a classic shortest path algorithm used to find the shortest path from a single source point to all other points in a weighted graph. In WSN, Dijkstra algorithm can determine the best path for data from ordinary nodes to cluster heads and then to base stations, thereby reducing network transmission delay and improving energy efficiency. First, the WSN node structure is clustered using an improved spectral clustering method. Secondly, the transmission path of the WSN node is preliminarily planned using the Dijkstra algorithm to construct the shortest path from each point to the cluster head. Both improved methods comprehensively consider factors such as node residual energy, distance, load, and number of times a relay node is selected. Combined with CRITIC to dynamically design the weight value of the edge, a final value is obtained to replace the self-edge weight of the undirected graph in Dijkstra and spectral clustering. According to the basic idea of ​​decision tree, the energy consumption values ​​of each Dijkstra algorithm transmission path and point-to-point transmission path are finally compared, and the path with the smallest energy consumption value is selected as the data transmission path.

[0008] To achieve the above object, the technical solution adopted by the present invention is: an optimized cluster routing design method applied in WSN, the specific steps are as follows:

[0009] Step 1: Calculate the residual energy factor E(n) between nodes in the overall environment ij , distance factor D(n) ij 、Load factor LF(n) ij , Number of elected cluster heads Freq(n) ij , the calculation methods are as follows:

[0010] 11) Residual energy factor E(n) ij :

[0011]

[0012] When nodes i and j are connected to each other, E(n) Connected-i Represents the residual energy factor of the i-th node, E(n) Connected-j Represents the residual energy factor of the jth node, E(n) ij Represents the mean residual energy of nodes i and j;

[0013]

[0014] Among them, E res (n)i represents the residual energy of the ith node, E res (n) j represents the remaining energy of the jth node, represents the average residual energy of the system and is calculated as follows:

[0015]

[0016] Among them, E res (q) represents the residual energy of the qth node, q represents the iteration variable, and the iteration range here is (1, N) where N is the number of nodes in the WSN;

[0017] 12) Distance factor D(n) ij :

[0018]

[0019] In the case where nodes i and j are connected to each other, X Connected-i represents the horizontal coordinate of node i, Y Connected-i represents the ordinate of node i, X Connected-j represents the horizontal coordinate of node j; Y Connected-j represents the ordinate of node j;

[0020] 13) Load factor LF(n) ij :

[0021]

[0022] In the case where nodes i and j are connected to each other, RP(n) i Represents the size of the data packet received by the i-th node, RP(n) j represents the size of the data packet received by the jth node, and TB is the total buffer size of the node;

[0023] 14) Number of elected cluster heads Freq(n) ij :

[0024]

[0025] Freq(n) i Indicates the number of times the i-th node is elected as the cluster head, Freq(n) j Indicates the number of times the jth node is elected as the cluster head, Freq(n) ij Represents Freq(n) i and Freq(n) j The average value of

[0026] Step 2: EWV ijIt represents the edge weight between the i-th point and the j-th point, and the weight expression is:

[0027] EWV ij =w 1 *D(n) ij +w 2 *Freq(n) ij +w 3 *E res (n) ij +w 4 *DLF(n) ij

[0028] Among them, w 1 Represents the distance factor between nodes D(n) ij The weight value, w 2 Freq(n) represents the average number of cluster heads elected among nodes ij The weight value, w 3 Represents the residual energy factor E between nodes res (n) ij The weight value, W 4 Indicates the load factor between nodes LF(n) ij The weight value is, according to the above formula It can be calculated.

[0029] Step 3: To calculate EWV ij , collect the indicator data D(n) between each node ij 、Freq(n) ij 、Eres(n) ij , LF(n) ij , normalize the collected indicator data to data of similar relative size and directly comparable, the steps are as follows:

[0030] 31) List the original indicator data matrix:

[0031]

[0032] The CRITIC weight method generally uses forward or reverse processing;

[0033] Positive indicators are:

[0034]

[0035] As a selection indicator, the larger the value of the positive indicator, the better.

[0036] Negative indicators are:

[0037]

[0038] As a selection indicator, the smaller the value of the negative indicator, the better.

[0039] If Y 11 It is a positive indicator. With the newly calculated Y 11 ' to replace the Y in the Y matrix 11 ) Among them, Y ij is the element in the i-th row and j-th column of the Y matrix, N is the number of nodes in the WSN network, and Y j-min represents the jth column in the Y matrix (i.e., the minimum value in the jth index), Y j-max Represents the jth column in the Y matrix (i.e. the maximum value in the jth index)

[0040] 32) Calculate the variability of indicators:

[0041]

[0042] Represents the average value of the jth indicator, through Y j The standard deviation of the jth indicator, i.e. the variability of the indicator, can be calculated:

[0043]

[0044] Among them, S j This is the variability of the indicator;

[0045] 33) Calculation index conflict:

[0046]

[0047] R j Indicates the conflict of indicators, r ij The calculation method of is as follows (using Pearson correlation coefficient):

[0048]

[0049] Among them, q represents the number of iterations of matrix column elements, FI qi represents the element in the ith column and qth row, SE qj represents the element in the jth column and the qth row, represents the mean value of the i-th column, represents the average value of the jth column;

[0050] 34) Calculate the amount of information:

[0051]

[0052] Among them, C j Indicates the amount of information; the greater the role of the j-th indicator in the entire indicator system, the more weight it should be assigned.

[0053] 35) Calculate objective weight:

[0054] The objective weight W of the jth indicator j for:

[0055]

[0056] So far, w has been calculated 1 、w 2 、w 3 、w 4 Calculate EWV according to Article 3 ij The EWV can be calculated by the formula ij .

[0057] Step 4: Perform spectral clustering operation:

[0058] 41) The similarity matrix Sm is calculated as follows:

[0059]

[0060] 42) The elements in the similarity matrix Sm are calculated as follows:

[0061]

[0062] Among them, W ij and W ji Both represent the weight between node i and node j;

[0063] 43), generate undirected spectral clustering;

[0064] 44) The degree matrix is ​​a diagonal matrix Dm consisting of N degrees, calculated as follows:

[0065]

[0066] In the data structure, the degree is defined as the number of vertices directly connected to the point, which is defined as follows:

[0067]

[0068] 45) Calculate the Laplace matrix L:

[0069] L=Dm-Sm

[0070] Perform eigendecomposition on the Laplace matrix, find the smallest non-zero eigenvalue and the corresponding eigenvector, and apply the hierarchical clustering algorithm to cluster in the new feature space;

[0071] To assess the quality of clustering, the Silhouette Coefficient or Davies-Bouldin can be used.

[0072] Step 5: Using EWV ij Instead of Dijkstra's edge weights, the shortest transmission path from each point to the cluster head is calculated within the cluster.

[0073] Compared with the prior art, the specific beneficial effects of the present invention are as follows: the algorithm proposed by the present invention aims to reduce the energy consumption of WSN when transmitting data in a large-scale detection area, improve the life cycle of nodes, reduce mortality, and enhance the energy balance of WSN networks. First, node clustering achieved by spectral clustering can balance energy consumption and avoid the formation of hot spots, thereby extending the life cycle of the entire network. Second, the Dijkstra algorithm can calculate the shortest path from an ordinary node to a cluster head and then to a base station, and the path value and weight are designed separately, which can effectively improve the efficiency of data transmission. Third, spectral clustering not only helps to identify clusters in the network, but also can realize local processing and aggregation of data within the cluster, reduce dependence on central nodes, realize load balancing, and help network administrators identify community structures in the network and effectively monitor and manage the network. Fourth, by combining spectral clustering and the Dijkstra algorithm and using CRITIC to calculate the path edge weights, the Dijkstra algorithm can be used within the cluster to find the optimal data forwarding path and reduce the transmission of redundant data. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 Schematic diagram of the initial WSN structure.

[0075] Figure 2 Schematic diagram of generating an undirected graph for spectral clustering.

[0076] Figure 3 This is a schematic diagram of the first round of spectral clustering results.

[0077] Figure 4 Flowchart for calculating the shortest transmission path from each point to the cluster head within the cluster.

[0078] Figure 5 This is a schematic diagram of the path from the final cluster members to the cluster head.

[0079] Figure 6 It is a life cycle comparison chart of the simulation experiment in the embodiment.

[0080] Figure 7 It is a comparison diagram of the residual energy of the simulation experiment in the embodiment. DETAILED DESCRIPTION

[0081] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0082] An optimized clustering routing design method applied in WSN uses an improved spectral clustering method to cluster the WSN node structure. In the clustered structure, an improved Dijsktra combined with the basic idea of ​​decision tree is used to select the optimal data transmission path.

[0083] The initial WSN structure is as follows Figure 1 The specific steps are as follows:

[0084] Step 1: Calculate the residual energy factor E(n) between nodes in the overall environment ij , distance factor D(n) ij 、Load factor LF(n) ij , Number of elected cluster heads Freq(n) ij , the calculation methods are as follows:

[0085] 11) Residual energy factor E(n) ij :

[0086]

[0087] When nodes i and j are connected to each other, E(n) Connected-i Represents the residual energy factor of the i-th node, E(n) Connected-j Represents the residual energy factor of the jth node, E(n) ij Represents the mean residual energy of nodes i and j;

[0088]

[0089] Among them, E res (n) i represents the residual energy of the ith node, E res (n) j represents the remaining energy of the jth node, represents the average residual energy of the system and is calculated as follows:

[0090]

[0091] Among them, E res (q) represents the residual energy of the qth node, q represents the iteration variable, and the iteration range here is (1, N) where N is the number of nodes in the WSN;

[0092] 12) Distance factor D(n) ij :

[0093]

[0094] In the case where nodes i and j are connected to each other, X Connected-i represents the horizontal coordinate of node i, Y Connected-i represents the ordinate of node i, X Connected-j represents the horizontal coordinate of node j; Y Connected-j represents the ordinate of node j;

[0095] 13) Load factor LF(n) ij :

[0096]

[0097] In the case where nodes i and j are connected to each other, RP(n) i Represents the size of the data packet received by the i-th node, RP(n) j represents the size of the data packet received by the jth node, and TB is the total buffer size of the node;

[0098] 14) Number of elected cluster heads Freq(n) ij :

[0099]

[0100] Freq(n) i Indicates the number of times the i-th node is elected as the cluster head, Freq(n) j Indicates the number of times the jth node is elected as the cluster head, Freq(n) ij Represents Freq(n) i and Freq(n) j The average value of

[0101] Step 2: EWV ij It represents the edge weight between the i-th point and the j-th point, and the weight expression is:

[0102] EWV ij =w 1 *D(n) ij +w 2 *Freq(n) ij +w 3 *E res (n) ij +w 4 *DLF(n) ij

[0103] Step 3: To calculate EWV ij, collect the indicator data D(n) between each node ij 、Freq(n) ij 、Eres(n) ij , LF(n) ij , normalize the collected indicator data to data of similar relative size and directly comparable, the steps are as follows:

[0104] 31) List the original indicator data matrix:

[0105]

[0106] The CRITIC weight method generally uses forward or reverse processing;

[0107] Positive indicators are:

[0108]

[0109] As a selection indicator, the larger the value of the positive indicator, the better.

[0110] Negative indicators are:

[0111]

[0112] As a selection indicator, the smaller the value of the negative indicator, the better.

[0113] If Y 11 It is a positive indicator. With the newly calculated Y 11 ' to replace the Y in the Y matrix 11 ; Among them, Y ij is the element in the i-th row and j-th column of the Y matrix, N is the number of nodes in the WSN network, and Y j-min represents the jth column in the Y matrix (i.e., the minimum value in the jth index), Y j-max Represents the jth column in the Y matrix (i.e. the maximum value in the jth index)

[0114] 32) Calculate the variability of indicators:

[0115]

[0116] Represents the average value of the jth indicator, through Y j The standard deviation of the jth indicator, i.e. the variability of the indicator, can be calculated:

[0117]

[0118] Among them, S j This is the variability of the indicator;

[0119] 33) Calculation index conflict:

[0120]

[0121] R j Indicates the conflict of indicators, r ij represents the correlation coefficient, r ij The calculation method of is as follows (using Pearson correlation coefficient):

[0122]

[0123] Among them, q represents the number of iterations of matrix column elements, FI qi represents the element in the ith column and qth row, SE qj represents the element in the jth column and the qth row, represents the mean value of the i-th column, represents the average value of the jth column;

[0124] 34) Calculate the amount of information:

[0125]

[0126] Among them, C j Indicates the amount of information; the greater the role of the j-th indicator in the entire indicator system, the more weight it should be assigned.

[0127] 35) Calculate objective weight:

[0128] The objective weight W of the jth indicator j for:

[0129]

[0130] So far, w has been calculated 1 、w 2 , W 3 、w 4 Calculate EWV according to step 2 ij The EWV can be calculated by the formula ij .

[0131] Step 4: Perform spectral clustering operation:

[0132] 41) The similarity matrix Sm is calculated as follows:

[0133]

[0134] 42) The elements in the similarity matrix Sm are calculated as follows:

[0135]

[0136] Among them, Wij and W ji Both represent the weight between i and j;

[0137] 43) Figure 2 As shown, undirected spectral clustering is generated;

[0138] 44) The degree matrix is ​​a diagonal matrix Dm consisting of N degrees, calculated as follows:

[0139]

[0140] In the data structure, the degree is defined as the number of vertices directly connected to the point, which is defined as follows:

[0141]

[0142] 45) Calculate the Laplace matrix L:

[0143] L=Dm-Sm

[0144] Perform eigendecomposition on the Laplace matrix, find the smallest non-zero eigenvalue and the corresponding eigenvector, and apply the hierarchical clustering algorithm to cluster in the new feature space.

[0145] To evaluate the quality of clustering, the silhouette coefficient or Davies-Bouldin can be used. The results of the first round of spectral clustering are as follows: Figure 3 shown.

[0146] Step 5: Using EWV ij Instead of Dijkstra's edge weight, the shortest transmission path from each point to the cluster head is calculated within the cluster. The specific process is as follows Figure 4 shown.

[0147] The final path from the cluster member to the cluster head is as follows: Figure 5 The specific steps are shown in the following table.

[0148]

[0149] In order to verify the algorithm mentioned in the present invention, the improved algorithm was simulated using PyCharmCommunityEdition2022.2.3 software. In the simulation experiment, a map of 100×100 was set, the unit was set to meter (m), the number of data nodes was set to 100, and iterations were performed respectively. The free space transmission attenuation coefficient was set to 1e-11, the multipath transmission attenuation coefficient was set to 1.3e-15, the energy consumption of sending / receiving data was set to 5e-8, and the energy consumption of fusing 1bit data was set to 5e-9J / bit. The specific parameter settings are as follows.

[0150]

[0151]

[0152] Based on the above parameters, the respective life cycle comparison is as follows Figure 6 shown.

[0153] When the first dead node appears, the number of rounds of the SC-DC algorithm is 525, while the number of rounds of the LEACH algorithm is 491. Compared with LEACH, the life cycle of SC-DC is increased by 6.92%. When half of the nodes die, the number of rounds of the SC-DC algorithm is 1121, while the number of rounds of the LEACH algorithm is 852. Compared with LEACH, the life cycle of SC-DC is increased by 31.57%. When all nodes die, the number of rounds of the SC-DC algorithm is 1642, while the number of rounds of the LEACH algorithm is 1084. Compared with LEACH, the life cycle of SC-DC is increased by 51.48%.

[0154] Based on the above parameters, the respective residual energy comparison is Figure 7 shown.

[0155] When the average system energy consumption reaches 0.25 (i.e. half), the number of rounds of the SC-DC algorithm is 592, while the number of rounds of the LEACH algorithm is 353. Compared with LEACH, the life cycle of SC-DC is increased by 67.71%. When the average system energy consumption reaches 0 (i.e. all dead), the number of rounds of the SC-DC algorithm is 1642, while the number of rounds of the LEACH algorithm is 1084. Compared with LEACH, the life cycle of SC-DC is increased by 51.48%.

[0156] The present invention reduces the energy consumption of WSN data transmission under the same time, increases the death cycle of WSN nodes, and enhances the energy balance of the WSN network.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. An optimized cluster routing design method applied in WSN, characterized in that: The specific steps are as follows: Step 1: Calculate the residual energy factor E(n) between nodes in the overall environment ij , distance factor D(n) ij 、Load factor LF(n) ij , Number of elected cluster heads Freq(n) ij ; Step 2: Calculate the edge weight EWV between the i-th point and the j-th point ij ; Step 3: Collect the indicator data D(n) between each node ij 、Freq(n) ij 、Eres(n) ij , LF(n) ij , normalize the collected indicator data, the specific steps are as follows: Step 31), list the original indicator data matrix as: Positive indicators are: Negative indicators are: Among them, Y ij is the element in the i-th row and j-th column of the Y matrix, N is the number of nodes in the WSN network, and Y j-min represents the minimum value of the jth index in the Y matrix, Y j-max Represents the maximum value of the jth index in the Y matrix; 32) Calculate the index variability as: Represents the average value of the jth indicator, through Y j Calculate the standard deviation of the jth indicator: Among them, S j Indicates the variability of the indicator; 33) Calculation index conflict: Among them, r ij represents the correlation coefficient and is calculated as follows: Among them, q represents the number of iterations of matrix column elements, FI qi represents the element in the ith column and qth row, SE qj represents the element in the jth column and the qth row, represents the mean value of the i-th column, represents the average value of the jth column; 34) Calculate the amount of information: 35) Calculate objective weight: The objective weight W of the jth indicator j for: Step 4: Perform spectral clustering operation: 41) The similarity matrix Sm is calculated as follows: 42) The elements in the similarity matrix Sm are calculated as follows: Among them, W ij and W ji Both represent the weight between node i and node j; 43), generate undirected spectral clustering; 44) The degree matrix is ​​a diagonal matrix Dm composed of N degrees, and the specific calculation is as follows: In the data structure, the degree is defined as the number of vertices directly connected to the point. The specific expression is as follows: 45) Calculate the Laplace matrix L: L=Dm-Sm Perform eigendecomposition on the Laplace matrix, find the smallest non-zero eigenvalue and the corresponding eigenvector, and apply the hierarchical clustering algorithm to cluster in the new feature space; Step 5: Using EWV ij The edge weights are calculated and the shortest transmission path from each point to the cluster head is calculated within the cluster.

2. The optimized cluster routing design method applied to WSN according to claim 1 is characterized in that: In step 1, the residual energy factor E(n) ij The specific calculation formula is: When nodes i and j are connected to each other, E(n) Connected-i Represents the residual energy factor of the i-th node, E(n) Connected-j Represents the residual energy factor of the jth node, E(n) ij Represents the mean residual energy of nodes i and j; Among them, E res (n) i represents the residual energy of the ith node, E res (n) j represents the remaining energy of the jth node, represents the average residual energy of the system, which is calculated as follows: Among them, E res (q) represents the residual energy of the qth node, q represents the iteration variable, and the iteration range here is (1, N) where N is the number of nodes in the WSN; Distance factor D(n) ij The specific calculation formula is: In the case where nodes i and j are connected to each other, X Connected-i represents the horizontal coordinate of node i, Y Connected-i represents the ordinate of node i, X Connected-j represents the horizontal coordinate of node j; Y Connected-j represents the ordinate of node j; Load factor LF(n) ij The specific calculation formula is: In the case where nodes i and j are connected to each other, RP(n) i Represents the size of the data packet received by the i-th node, RP(n) j represents the size of the data packet received by the jth node, TB represents the total buffer size of the node; Number of elected cluster heads Freq(n) ij The specific calculation formula is: Freq(n) i Indicates the number of times the i-th node is elected as the cluster head, Freq(n) j Indicates the number of times the jth node is elected as the cluster head, Freq(n) ij Represents Freq(n) i and Freq(n) j The average value of .

3. The optimized cluster routing design method applied to WSN according to claim 2 is characterized in that: In step 2, the weight EWV ij The specific expression is: EWV ij =w1*D(n) ij +w2*Freq(n) ij +w3*E res (n)ij+w4*LF(n) ij Where w1 represents the distance factor D(n) between nodes ij The weight value of w2 represents the average factor of the number of cluster heads elected among nodes Freq(n) ij The weight value, w3 represents the residual energy factor E between nodes res (n) ij The weight value, w4 represents the load factor LF(n) between nodes ij The weight value of .

Images