A high-efficiency clustering method for heterogeneous wireless sensor networks based on genetic algorithm

Abstract

The present invention proposes a high-energy-efficiency clustering method for heterogeneous wireless sensor networks based on genetic algorithms. This method continues the idea of ​​genetic algorithms, converts the exploration problem of clustering strategies into the update problem of coding chromosomes, searches and screens to find the chromosomes with the best performance, maps them to the clustering strategies in reality, and finally feeds back to the network for strategy execution. In the search process, the crossover operation is used to accelerate the search for the optimal solution of the strategy, and the mutation operation is used to avoid falling into the local optimum during the search process; in the screening process, chromosomes with excellent performance are screened according to the fitness value, and the fitness function for evaluating the quality of chromosomes is designed by comprehensively considering factors such as the energy consumption of single-round network operation, the load of cluster head nodes, and the communication scheduling delay.

Description

Technical Field

[0001] The invention belongs to the field of wireless sensor networks, and in particular relates to a node clustering method for heterogeneous wireless sensor networks. More specifically, the invention uses a high-energy-efficiency clustering method for heterogeneous wireless sensor networks improved based on a genetic algorithm (GA). Background Art

[0002] With the increasing demand for digital twin modeling of multimodal spatiotemporal information in different fields in the construction of smart cities and new infrastructure, the application scenarios of deploying large-scale sensors as information acquisition equipment in different spaces such as the surface, low altitude, indoors, and underground are gradually increasing. Sensors acquire multimodal spatiotemporal perception data such as high-resolution images, videos, point clouds, and environmental parameters. With the technical support of wireless communication and satellite Internet, the perception data is forwarded through the base station (BS), and finally transmitted to different types of services to control terminals such as cloud-based load task managers to achieve real-time reconstruction of scenes and dynamic information interaction feedback linkage response. In such application scenarios, the deployed sensors are usually of various types, and there are differences in computing, perception, and energy storage performance among different types, making sensor energy heterogeneity a common phenomenon, thus forming heterogeneous wireless sensor networks (HWSNs).

[0003] Factors such as size and load limit the energy reserve of sensors, and energy replenishment is subject to environmental constraints of application scenarios, which makes HWSNs have higher requirements for extending the lifetime (LT) of all nodes in the network, and pays more attention to how to reduce the cost of the communication link from the collection area to the remote end. In addition, the spatiotemporal correlation between nodes causes a large amount of redundant data in the network, and environmental noise and node failures affect the reliability of perception data. When large-scale sensors are clustered, the cluster head (CH) node uses the data fusion data processing technology advantage to analyze, summarize and organize the perception data collected by cluster members (CM), while ensuring the accuracy of detection information and reducing the data traffic in the network. Therefore, seeking high-energy-efficiency clustering methods with non-deterministic polynomial-time hard (NP-Hard) characteristics with polynomial complexity has become one of the important research directions of HWSNs.

[0004] For solving NP-hard problems, the main methods can be divided into heuristic algorithms (HA) and metaheuristic algorithms (MA). HA can change the search path according to individual or global experience. It is a method for efficiently obtaining feasible solutions, but it lacks stability and is prone to fall into local optimality. As an improvement of HA, MA uses random search techniques to explore the solution space, which is more suitable for optimization problems when information is incomplete or computing power is limited. Therefore, the present invention integrates the MA idea into network clustering and proposes a high-energy-efficiency clustering method based on genetic algorithm. Combined with the energy heterogeneity characteristics of HWSNs, a fitness function weighted by multiple factors such as sensor residual energy, distribution location and perceived data volume is designed. The optimal cluster selection strategy with low energy consumption is found through GA optimization. While ensuring the perceived data volume and communication service quality (Quality of Service, QoS), the network LT is extended and the energy utilization rate of the entire network is improved. Summary of the invention

[0005] The purpose of the present invention is to address the problems of short node life and unbalanced load between clusters in HWSNs, and propose a high-efficiency clustering method for heterogeneous wireless sensor networks based on genetic algorithm, which can extend the network life while ensuring the amount of perceived data and QoS. In order to achieve this purpose, the steps adopted by the present invention are:

[0006] Step 1: Build an energy heterogeneous wireless sensor network scenario, which consists of 1 base station node and N wireless sensor nodes. According to different initial energy settings, the sensor nodes are subdivided into advanced nodes and ordinary nodes. The initial energy of advanced nodes is higher than that of ordinary nodes.

[0007] Step 2: The position and remaining energy of each node in the above network scenario are used as the input information of the genetic algorithm, and the clustering strategy is mapped into the coding chromosome. Through single-point exchange and local mutation, the population diversity is enriched and the population iteration is realized;

[0008] Step 3: According to the fitness function of the clustering strategy, the fitness values ​​of the initial chromosome, crossover offspring and mutant offspring are calculated;

[0009] Step 4: Arrange the chromosomes in ascending order according to the fitness values, determine the new population generated by a single round of iteration, and use it as the initial population for the next round of iteration;

[0010] Step 5: According to the results of the last round of population iteration, determine the chromosome with the smallest fitness value and convert it into the optimal clustering strategy;

[0011] Step 6: Clustering is performed according to the optimal strategy searched by the genetic algorithm. Cluster members transmit the perception information to their cluster head node in a single hop. The cluster head performs data fusion and sends the information to the base station.

[0012] The high-efficiency clustering method for heterogeneous wireless sensor networks based on genetic algorithm proposed in this invention has been implemented in Matlab R2017b simulation software. In order to simulate different network environments, Table 1 lists the abbreviations of the parameters related to the network scenario construction, and the values ​​of the remaining simulation parameters are shown in Table 2.

[0013] Table 1 Abbreviations of simulation scenario parameters

[0014]

[0015] Table 2 Simulation parameters

[0016]

[0017] Set the parameters of the simulation scenario [λ, E nom , E adv , P] is [20%, 2J, 4J, 0.1], the length of the simulation area is 200m, the number of deployed sensors is 100, and the base station coordinates are (100m, 200m). The performance simulation results of the present invention are compared with those of the traditional heuristic clustering methods LEACH, MHCRA and other genetic clustering methods GAECH. Figure 3 A comparison chart of the number of remaining surviving nodes in the network changing with simulation time under four clustering methods is given. Figure 4 The comparison of the last node dead (LND) rounds of each clustering method under different area scales (L = 100m, 200m, 300m) is given. Appendix 1 shows the comparison of the number of LND rounds and the amount of perception data packets sent by each clustering method when different numbers of nodes are deployed (N = 100, 200, 300). It can be seen that the present invention has performance advantages over the other three clustering methods in improving the overall energy utilization of the network and extending the network lifetime. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the mapping from iterative population to clustering solution space;

[0019] Figure 2 A flow chart of chromosome conversion and strategy execution provided by an embodiment of the present invention;

[0020] Figure 3 This is a comparison chart of the number of remaining surviving nodes of the four clustering methods as the number of simulation rounds changes;

[0021] Figure 4This is the LND round change diagram of the network under different regional scales;

[0022] Table 1 shows the LND rounds / sensing data packet sending volume under different node scales. DETAILED DESCRIPTION

[0023] The core idea of ​​the present invention is: in the face of the problem of obtaining a high-energy-efficiency clustering strategy with NP-hard characteristics, the present invention adopts the idea of ​​seeking the optimal solution using a genetic algorithm, converts the exploration problem of the clustering strategy into the update problem of the coding chromosome, finds the chromosome with the best performance through the two processes of searching and screening, and then maps it to the clustering strategy in reality, and finally feeds it back to the network for strategy execution. In the search process, the crossover operation is used to accelerate the search for the optimal solution of the strategy, and the mutation operation is used to avoid falling into the local optimum during the search process. The above two operations on chromosomes are used to achieve population iteration, and then explore the solution space; in the screening process, chromosomes with excellent performance are screened according to the fitness value, and factors such as single-round network operation energy consumption, cluster head node load and communication scheduling delay are comprehensively considered, and the sub-function weights are optimized in combination with the magnitude difference of the objective function, and a fitness function for evaluating the quality of chromosomes is designed.

[0024] The specific implementation steps of the present invention are given below:

[0025] Step 1: Build network scenarios and communication models.

[0026] The simulated HWSNs scenario consists of a base station node and N wireless sensor nodes. All sensor nodes are randomly deployed in the square simulation area. After deployment, the nodes remain in the same position. According to the initial energy, the sensor nodes are subdivided into advanced nodes and ordinary nodes. The initial energy set for the former is higher than that for the latter. The communication module of the sensor has four states: sending, receiving, idle and sleeping. The process of sending and receiving data consumes the main energy. The energy ET consumed by sending 1 bit of data is calculated as follows:

[0027]

[0028] Where d represents the Euclidean distance between the transmitting and receiving ends; E elec Represents the energy consumed by the transmitting circuit to process each bit of data; ε fs is the amplification factor of the free space propagation model; ε mp is the amplification factor of the multipath fading channel model; the distance threshold d 0 for:

[0029]

[0030] Different from the direct reception of control packets such as broadcasts and requests, the node serving as the cluster head also needs to use the stickiness between data to filter redundant information and compress the length of the sensed data to the fusion rate R agg The weighted number of effective information bits. Therefore, according to the type of received packets, the energy consumption formula when the sensor receives 1 bit of data is defined as:

[0031]

[0032] Where E agg It represents the energy consumption of cluster head node to fuse unit bit of data.

[0033] Step 2: Use the node position and energy information of the simulation area as the input of the genetic algorithm, map the clustering strategy to the encoding chromosome, enrich the diversity of chromosomes in the population through crossover and mutation operations, and realize population iteration.

[0034] The population iteration and mapping process is as follows Figure 1 As shown, the following steps are included:

[0035] Step 2-1: Random election of cluster heads. Sensors with residual energy lower than the energy threshold δ are considered dead nodes and will not participate in clustering. Binary code the surviving nodes and take a random number uniformly distributed in the interval [0, 1). If the random number is greater than the set initial cluster head ratio P, the node is marked as "1"; otherwise, it is marked as "0", thus obtaining the N alive The initial chromosome consists of surviving nodes;

[0036] Step 2-2: Get N by repeating step 2-1 pop chromosomes to construct the initial population;

[0037] Step 2-3: Select chromosomes with a pc ratio from the initial population for single-point crossover operation. The length of the chromosome depends on the number of surviving nodes N in the current network state. alive , in 1 to N alive -1, the parent chromosomes exchange some genetic information, and the following two daughter chromosomes are generated:

[0038]

[0039]

[0040] Step 2-4: Select chromosomes with a ratio of pm from the initial population for mutation. In the selected parent chromosome, randomly select the tth gene for local mutation, where the integer t is randomly selected from 1 to N. alive -1, the resulting mutant progeny chromosomes are:

[0041]

[0042] Step 2-5: Map the algorithm search space to the problem solution space, that is, map the binary coded chromosome to the actual clustering strategy. Figure 2 As shown in the figure, if there is no "1" in the chromosome, that is, the cluster head set in this strategy is empty, then all nodes will send the perception information directly to the base station. The node marked as "0" is used as a cluster member and pre-joins the "1" node that is closest to it. If the "1" node has a "0" node pre-joining it, that is, the cluster member set of the cluster is not empty, then the node is elected as the cluster head, otherwise it becomes a cluster member. In this way, the chromosome encoding value is mapped to the identity of each surviving node, and the topological structure of the cluster is determined.

[0043] Step 3: Considering multiple factors such as single-round energy consumption, data volume, latency, and residual energy, the fitness values ​​of the initial chromosomes, crossover offspring, and mutant offspring are calculated.

[0044] The fitness function design process includes the following steps:

[0045] Step 3-1: The energy consumption generated by a single round of data transmission and reception is a direct factor affecting the current network energy. The total energy consumption includes the cluster head broadcast energy consumption E broad , cluster entry energy consumption E join And the energy consumption of perceptual information transmission E data . Define the subfunction f 1 for:

[0046] f 1 =E broad +E join +E data (7)

[0047]

[0048]

[0049]

[0050] Among them C num represents the number of clusters, CM∈C k Indicates that the cluster member belongs to the kth cluster, d broad Represents the broadcast radius, R agg represents the fusion rate of each bit of data by the cluster head, L ctr Indicates the length of the broadcast / request / response control data packet, L data It represents the length of the sensing data packet sent by each cluster member in a single round.

[0051] Step 3-2: Considering only the energy consumption of the entire network, when all sensor nodes act as cluster heads, the network energy consumption is the lowest. At this time, the strategy has lost the essential function of the sensor network. To a certain extent, reducing the proportion of cluster heads can exchange for more sensor data to be collected by cluster members. Therefore, the sub-function f 2 Designed for:

[0052]

[0053] Where N alive Indicates the number of live nodes in the current HWSNs.

[0054] Step 3-3: Compared with cluster members that can be in a sleep state with low energy consumption, cluster members need more energy storage to achieve idle monitoring, data transmission and reception, and fusion functions. An excellent clustering strategy should tend to elect nodes with more remaining energy as cluster heads. Therefore, the sub-function f 3 Defined as:

[0055]

[0056] Where E CHk and E CM They represent the residual energy of the kth cluster head and its members respectively.

[0057] Step 3-4: Use the statistics of the distance from the cluster head to the base station as the sub-function f 4 :

[0058]

[0059] Step 3-5: Since the communication adopts time division multiple access, if there are too many members in the cluster, the cluster head needs more time slots to allocate channels for the members, which will affect the real-time performance of data transmission. Therefore, define the sub-function f 5 for:

[0060]

[0061] Where N Ck Indicates the number of cluster members owned by the kth cluster.

[0062] Step 3-6: Use subfunction f 6 Measure the deviation of the total residual energy between clusters:

[0063]

[0064] Step 3-7: Node spacing can be used as one of the indicators for evaluating the node dispersion in the topological structure. If the minimum distance between cluster heads is larger, it means that the inter-cluster dispersion is higher and the inter-cluster interference is smaller; if the maximum distance between cluster members and cluster heads is smaller, it means that the cluster members are closer to the cluster heads and the energy consumption of intra-cluster transmission and reception is lower. Therefore, this method designs a sub-function f7 :

[0065]

[0066] Step 3-8: Considering the above single-round energy consumption, data volume, latency, and remaining energy, a fitness function fitness is designed. The smaller the value, the better the clustering strategy.

[0067]

[0068] where ω i Represents the weight coefficient, f i Represents each sub-function.

[0069] Step 4: Arrange the chromosomes in ascending order according to the fitness value and select the first N pop The chromosomes are used to construct a new population after a single round of iteration, which is used as the initial population for the next round of iteration.

[0070] Step 5: According to the results of the last round of population iteration, determine the chromosome with the smallest fitness value and convert it into the optimal clustering strategy. This conversion process is consistent with steps 2-5;

[0071] Step 6: Clustering is performed according to the optimal strategy searched by the genetic algorithm. Cluster members transmit the perception information to their cluster head node in a single hop. The cluster head performs data fusion and sends the information to the base station.

[0072] The contents not described in detail in the present application belong to the prior art known to the professional and technical personnel in this field.

Claims

1. A high-efficiency clustering method for heterogeneous wireless sensor networks based on genetic algorithm, characterized in that: The steps include: Step 1: Generate binary coded chromosomes based on random election method and construct N pop The initial population consists of chromosomes, and the population iteration is achieved through the operations of single-point crossover and local mutation; Step 2: According to the designed fitness function, obtain the fitness value of the chromosome; Step 3: Arrange the chromosomes in ascending order according to the fitness values, determine the new population generated by a single round of iteration, and use it as the initial population for the next round of iteration; Step 4: According to the results of the last round of population iteration, the chromosome with the smallest fitness value is obtained, mapped to the optimal clustering strategy, and the node topology structure is determined; Step 5: Execute the optimal clustering strategy. The cluster member nodes transmit the perception information to their cluster head node in a single hop. The cluster head node performs data fusion and sends the information to the base station. Furthermore, in step 2, the fitness function design for evaluating chromosome performance includes the following specific steps: Step 2-1: Use sub-function f1 to calculate the energy consumption of a single round of data transmission and reception, including cluster head broadcast energy consumption E broad , cluster entry energy consumption E join And the energy consumption of transmitting perception information E data : f1=E broad +E join +E data Step 2-2: The sensors with residual energy higher than the energy threshold δ are determined as surviving nodes, and the proportion of cluster head (CH) nodes to the total surviving nodes is counted, which is designed as a sub-function f2: Among them C num Represents the number of clusters, N alive Indicates the number of surviving nodes in the current network; Step 2-3: By calculating the sub-function f3, the ratio of the average cluster member residual energy to the average cluster head residual energy is obtained: where CM∈C k Indicates that the cluster member belongs to the kth cluster, and E CM They represent the residual energy of the kth cluster head and its members respectively; Step 2-4: Use sub-function f4 to calculate the sum of the distances from the cluster head to the base station: Step 2-5: Count the clusters with the most cluster members, and use the ratio of the number of cluster members to the total surviving nodes as sub-function f5: in Indicates the number of cluster members owned by the kth cluster; Step 2-6: Use sub-function f6 to count the deviation of residual energy between clusters: Step 2-7: Calculate the ratio of the maximum distance between the cluster head and the cluster members to the minimum distance between the cluster heads through sub-function f7: Step 2-8: Combine the sub-functions from step 2-1 to step 2-7 to get the fitness function fitness: where ω i Represents the weight coefficient, the weight sum is 1, f i Represents a sub-function; Wherein, the step 4 includes the following specific steps: Step 4-1: Determine whether there is a mark "1" in the binary coding chromosome. If there is no mark "1", all nodes will send the sensing information directly to the base station and jump to step 5; otherwise, go to step 4-2; Step 4-2: Take the "0" marked node as a cluster member and pre-join the "1" marked node closest to it; Step 4-3: If a "1" marked node has a "0" node pre-joined, the "1" marked node is elected as the cluster head; otherwise, the "1" marked node becomes a cluster member.

Images