Efficient reliability-oriented super-large-scale wireless sensor network energy-saving deployment optimization method

Abstract

The invention relates to an efficient reliability-oriented super-large-scale wireless sensor network energy-saving deployment optimization method. The method comprises the following steps: firstly, obtaining a feature set of sensor nodes in each sub-region; obtaining a deployment scheme of each sensor node through forward propagation of the neural network, and converting optimization of the deployment scheme of the sensor nodes with high-dimensional decision variables into optimization of a low-dimensional weight matrix of the neural network through the neural network; and optimizing the neural network low-dimensional weight matrix by adopting a meta-heuristic algorithm to obtain a new neural network low-dimensional weight matrix and a deployment scheme of each sensor node. Through the setting, the region to be deployed is divided into a plurality of sub-regions, parallel processing is carried out, the deployment requirement of each sub-region is ensured, and a complex node deployment optimization problem is converted into optimization of a low-dimensional weight matrix. In addition, when C-connection is not satisfied among the sub-regions, a repair region is determined by constructing an adjacent set and selecting key nodes, additional nodes are deployed by using a meta-heuristic algorithm, and the overall C-connection of the network is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of energy-saving deployment of wireless sensor networks, and in particular to an efficient reliability-oriented energy-saving deployment optimization method for ultra-large-scale wireless sensor networks. Background Art

[0002] Wireless sensor networks are composed of battery-powered and easy-to-deploy sensor nodes with a self-organizing architecture that can monitor environmental parameters and transmit data to terminals through multiple hops. Ultra-large-scale wireless sensor networks contain a large number of energy-constrained nodes and are widely used in large farms, virgin forests and other scenarios. In the study of wireless sensor networks, target coverage and network connectivity are the core issues. Unreasonable node deployment will cause coverage holes, and node failure and battery exhaustion will also affect the reliability of target coverage. The introduction of K-coverage and Q-coverage has solved the related problems to a certain extent. The lack of network connectivity will make the data transmission link disconnected and hinder information aggregation. C-connectivity can alleviate such problems. In addition, the sensor node deployment strategy has a significant impact on coverage, connectivity and energy consumption. When deploying large-scale networks, there is a key problem that needs to be solved urgently - the MinEQC problem, that is, how to take into account heterogeneous sensor nodes, target Q coverage and node C-connectivity, and achieve energy saving on this basis. And how to solve the MinEQC problem efficiently and quickly is still a key challenge facing the current field of wireless sensor networks. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide an efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method, the method comprising:

[0004] Obtaining the area to be deployed, dividing the area to be deployed into multiple independent sub-areas and optimizing each sub-area in parallel, and obtaining a feature set of sensor nodes in each sub-area;

[0005] According to the feature set of sensor nodes in each sub-area, the deployment scheme of each sensor node is obtained by using the forward propagation of the neural network; and the neural network converts the optimization of the sensor node deployment scheme with high-dimensional decision variables into the optimization of the low-dimensional weight matrix of the neural network through dimensionality reduction processing;

[0006] A meta-heuristic algorithm and a fitness function are used to optimize the low-dimensional weight matrix of the neural network, and the meta-heuristic algorithm obtains a deployment plan of new sensor nodes according to the optimized low-dimensional weight matrix of the neural network;

[0007] It is determined whether each sub-region satisfies C-connectivity. When each sub-region does not satisfy C-connectivity, the meta-heuristic algorithm deploys additional sensor nodes to ensure that each sub-region satisfies C-connectivity.

[0008] Preferably, a meta-heuristic algorithm and a fitness function are used to optimize the low-dimensional weight matrix of the neural network, specifically including: the fitness function calculates the deployment plan of each sensor node and obtains the fitness value, and the meta-heuristic algorithm obtains a new low-dimensional weight matrix of the neural network and the deployment plan of each sensor node according to the fitness value.

[0009] Preferably, when the number of evaluations of the fitness function is less than or equal to the set total number of evaluations, the fitness function is used to calculate the deployment scheme of each sensor node and obtain a fitness value, and the meta-heuristic algorithm obtains a new low-dimensional weight matrix of the neural network and a deployment scheme of each sensor node according to the fitness value;

[0010] When the evaluation times of the fitness function is greater than the set total evaluation times, the deployment plans of the sensor nodes in the multiple sub-areas are merged to obtain the deployment plan of the entire area to be deployed.

[0011] Preferably, the dimension reduction process of the neural network also includes Gaussian perturbation and activation function; the Gaussian perturbation is used to prevent the characteristics of different types of sensor nodes in the same potential position from converging; the activation function converts the linear output of neurons in the neural network into nonlinear output by processing the input data of neurons in the hidden layer and output layer.

[0012] Preferably, the activation function processes the input data of neurons in the hidden layer and the output layer as follows:

[0013] For neurons in the hidden layer, the activation function formula is as follows:

[0014]

[0015] Where α = 0.01;

[0016] For the neurons in the output layer, the activation function formula is as follows:

[0017]

[0018] Preferably, the specific steps of judging whether C-connectivity is satisfied between each sub-region include: determining an adjacency set consisting of a plurality of adjacent sub-regions, and selecting a key sensor node and the coordinates of the key sensor node from the adjacent sub-regions constituting the adjacency set; forming a repair area according to the coordinates of the key sensor node; and using the judgment result of whether all sensor nodes in the repair area satisfy C-connectivity as the judgment result of whether C-connectivity is satisfied between each sub-region.

[0019] Preferably, the key sensor node has the shortest Euclidean distance to the geometric center of the adjacency set.

[0020] Preferably, when the number of sub-regions is sufficient, the adjacent set consists of four adjacent sub-regions; when the number of sub-regions is limited, the adjacent set consists of two adjacent sub-regions.

[0021] Preferably, the area to be deployed is divided according to the X and Y axes respectively to obtain sub-regions, the number of adjacent sets according to different X and Y values As shown below:

[0022]

[0023] in Represents round down, Z + Represents a non-zero natural number.

[0024] Preferably, the feature set of the sensor node includes the coverage radius of the sensor node, the coverage energy consumption of the sensor node, and the coordinates of the sensor node.

[0025] Beneficial effects of the present invention: The present invention divides the area to be deployed into multiple sub-areas, and adopts a parallel processing solution to solve each sub-area through neural network dimensionality reduction and meta-heuristic algorithm. The complex node deployment scheme optimization problem is converted into the optimization of a low-dimensional matrix, which greatly reduces the amount of calculation and search space, speeds up the solution, and enables the algorithm to quickly find a better node deployment plan. This method can avoid neural network training, does not require prior knowledge, and improves the solution efficiency. At the same time, by constructing an adjacency set, selecting key nodes to determine the repair area, and using a meta-heuristic algorithm to deploy additional nodes to repair connectivity. This process is highly targeted and can efficiently solve the connectivity problem between sub-areas, ensuring the C-connectivity of the entire network. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.

[0027] Figure 1 It is a flowchart of the present invention;

[0028] Figure 2 A detailed flow chart of a two-stage parallel framework for solving the MinEQC problem in a large-scale wireless sensor network in an embodiment of the present invention;

[0029] Figure 3 is the total energy consumption graph corresponding to all combinations of neural networks and metaheuristic algorithms involved in the present invention when δ=1.0 and different combinations of Q and C;

[0030] Figure 4 The constraint violation values ​​of the infeasible solutions obtained by the PSO and GA involved in the present invention when δ=1.0 and different combinations of Q and C;

[0031] Figure 5 is a ratio diagram of feasible solutions obtained by PSO, GA and all neural network combinations involved in the present invention when δ=1.0 and different combinations of Q and C;

[0032] Figure 6 is a graph of the total running time of all neural network and metaheuristic algorithm combinations involved in the present invention when δ=1.0 and different combinations of Q and C;

[0033] Figure 7 is a total energy consumption diagram of the MinEQC problem of the combination of the RNN and the meta-heuristic algorithm involved in the present invention when δ=0.5 and different combinations of Q and C;

[0034] Figure 8 is the total energy consumption diagram of the combination of RNN and ACO, CSO, and AROA involved in the present invention when δ=0 and different combinations of Q and C;

[0035] Fig. 9 is the constraint violation value of the infeasible solution of the combination of RNN and metaheuristic algorithm involved in the present invention when δ=0 and different combinations of Q and C;

[0036] Among them, RNN is a recurrent neural network, PSO is a particle swarm optimization algorithm, ACO is an ant colony optimization algorithm, GA is a genetic algorithm, CSO is a competitive particle swarm algorithm, and AROA is an attraction-repulsion optimization algorithm. DETAILED DESCRIPTION

[0037] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.

[0038] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two non-identical entities with the same name or non-identical parameters. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. The subsequent embodiments will not explain this one by one.

[0039] The directions and positions mentioned in the present invention, such as "upper", "lower", "front", "back", "left", "right", "inside", "outside", "top", "bottom", "side", etc., are only for reference to the directions or positions of the drawings. Therefore, the directions and positions used are for explaining and understanding the present invention, but not for limiting the scope of protection of the present invention.

[0040] like Figures 1 to 9 As shown in the figure, in an embodiment of the present invention, an efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method is provided, and the method includes:

[0041] Obtaining the area to be deployed, dividing the area to be deployed into multiple independent sub-areas and optimizing each sub-area in parallel, and obtaining a feature set of sensor nodes in each sub-area;

[0042] According to the feature set of sensor nodes in each sub-area, the deployment scheme of each sensor node is obtained by using the forward propagation of the neural network; and the neural network converts the optimization of the sensor node deployment scheme with high-dimensional decision variables into the optimization of the low-dimensional weight matrix of the neural network through dimensionality reduction processing;

[0043] A meta-heuristic algorithm and a fitness function are used to optimize the low-dimensional weight matrix of the neural network, and the meta-heuristic algorithm obtains a deployment plan of new sensor nodes according to the optimized low-dimensional weight matrix of the neural network;

[0044] It is determined whether each sub-region satisfies C-connectivity. When each sub-region does not satisfy C-connectivity, the meta-heuristic algorithm deploys additional sensor nodes to ensure that each sub-region satisfies C-connectivity.

[0045] A meta-heuristic algorithm and a fitness function are used to optimize the low-dimensional weight matrix of a neural network, specifically comprising: the fitness function calculates the deployment plan of each sensor node and obtains a fitness value, and the meta-heuristic algorithm obtains a new low-dimensional weight matrix of the neural network and a deployment plan of each sensor node according to the fitness value.

[0046] When the number of evaluations of the fitness function is less than or equal to the set total number of evaluations, the fitness function is used to calculate the deployment plan of each sensor node and obtain a fitness value, and the meta-heuristic algorithm obtains a new low-dimensional weight matrix of the neural network and a deployment plan of each sensor node according to the fitness value;

[0047] When the evaluation times of the fitness function is greater than the set total evaluation times, the deployment plans of the sensor nodes in the multiple sub-areas are merged to obtain the deployment plan of the entire area to be deployed.

[0048] The dimension reduction process of the neural network also includes Gaussian perturbation and activation function; the Gaussian perturbation is used to prevent the characteristics of different types of sensor nodes at the same potential position from converging; the activation function converts the linear output of neurons in the neural network into nonlinear output by processing the input data of neurons in the hidden layer and output layer.

[0049] The activation function processes the input data of the neurons in the hidden layer and the output layer as follows:

[0050] For neurons in the hidden layer, the activation function formula is as follows:

[0051]

[0052] Where α = 0.01;

[0053] For the neurons in the output layer, the activation function formula is as follows:

[0054]

[0055] The specific steps of judging whether C-connectivity is satisfied between each sub-region include: determining an adjacency set consisting of a plurality of adjacent sub-regions, and selecting a key sensor node and the coordinates of the key sensor node from the adjacent sub-regions constituting the adjacency set; forming a repair area according to the coordinates of the key sensor node; and using the judgment result of whether all sensor nodes in the repair area satisfy C-connectivity as the judgment result of whether C-connectivity is satisfied between each sub-region.

[0056] The key sensor node has the shortest Euclidean distance to the geometric center of the adjacency set.

[0057] When the number of subregions is sufficient, the adjacent set consists of four adjacent subregions; when the number of subregions is limited, the adjacent set consists of two adjacent subregions.

[0058] It can be understood that the shortest Euclidean distance between the key sensor node and the geometric center of the adjacency set is to minimize the repair area in the adjacency set; when the number of sub-areas is sufficient, the adjacency set consists of four adjacent sub-areas; when the number of sub-areas is limited, the adjacency set consists of two adjacent sub-areas, which is to minimize the number of adjacency sets.

[0059] Divide the area to be deployed by X and Y axes respectively, and get sub-regions, the number of adjacent sets according to different X and Y values As shown below:

[0060]

[0061] in Represents round down, Z + Represents a non-zero natural number.

[0062] The feature set of the sensor node includes the coverage radius of the sensor node, the coverage energy consumption of the sensor node, and the coordinates of the sensor node.

[0063] In the embodiment of the present invention, the main workflow is divided into two stages. The first is stage one. The work of stage two is to use neural networks and meta-heuristic algorithms to obtain a deployment plan for the entire area to be deployed before judging whether the sub-areas meet C-connectivity. Next, the C-connectivity between the sub-areas is judged. If the sub-areas meet C-connectivity, the workflow of this embodiment ends; if the sub-areas do not meet C-connectivity, enter stage two. The work of stage two is to deploy additional sensor nodes through a meta-heuristic algorithm so that the sub-areas meet C-connectivity.

[0064] The following is a description of the MinEQC problem:

[0065] 1. Overlay Model

[0066] In a deterministic deployment environment, the potential deployment location of each sensor node is known in advance. represents the potential deployment location of sensor nodes, where represents the total number of potential deployment locations, represents the coordinates of potential deployment locations. represents the total number of types of heterogeneous sensor nodes, Indicates whether the sensor node of type v is deployed at the potential location l i If deployed, then otherwise,

[0067] For all potential deployment locations, the deployment strategy can be described as follows:

[0068]

[0069] in A potential location i At most one sensor node is deployed, i.e.

[0070] make represents all target sets that need to be covered, where represents the total number of targets, represents the coordinates of the target. For heterogeneous sensor nodes, let represents the coverage radius of v type sensor nodes deployed at potential locations l iThe v-type sensor node has a j The coverage status can be described as follows:

[0071]

[0072] Where 1 means the target is covered, 0 means the target is not covered, and ||·|| means l i and t j The Euclidean distance between .

[0073] For the target t j , the total number of sensor nodes that can cover it is described as follows:

[0074]

[0075] make To describe the target j Whether it is covered by at least one sensor node, which is described as follows:

[0076]

[0077] Target Set The overall coverage is described as follows:

[0078]

[0079] 2 Communication model

[0080] For deployment in potential locations i and l j The communication probability between two sensor nodes is described as follows:

[0081]

[0082] Where R c represents the communication radius of the sensor node, R p ,λ,β,are both uncertainty measures that represent the communication model. φ(l i ,l j ) indicates deployment at potential location l i and l j The communication state between the two sensor nodes is controlled by the threshold ε, which is described as follows:

[0083]

[0084] For deployment in l i The total number of sensor nodes that it can directly communicate with is described as follows:

[0085]

[0086] The link for data transmission between sensor nodes can be abstracted into an undirected graph G wsn , the depth-first traversal strategy dft(·) is used to determine whether it is a connected graph. wsn is a connected graph, then dft(G wsn )=1. Otherwise, dft(G wsn )=0.

[0087] 3 Energy consumption model

[0088] Let e s represents the coverage energy consumption of heterogeneous sensor nodes, which is described as follows:

[0089]

[0090] Let e c represents the communication energy consumption of heterogeneous sensor nodes, which is described as follows:

[0091]

[0092] The total coverage energy consumption of the sensor nodes is calculated as follows:

[0093]

[0094] The total communication energy consumption of the sensor nodes is calculated as follows:

[0095]

[0096] Therefore, the energy consumption in phase 1 is described as follows:

[0097]

[0098] The C1 constraint ensures C-connectivity of all sensor nodes. C2 indicates that at most one sensor node is deployed in a potential location. C3 ensures that the data link graph between all sensor nodes is a connected graph. C4 ensures that the targets meet Q-coverage. C5 ensures that all targets achieve 100% coverage. Due to the hardware constraints of sensor nodes, such as memory and processor, C6 ensures that the maximum number of targets that can be covered by a sensor node is C7 ensures that a sensor node can communicate with a maximum of

[0099] The constraint processing method adopts the penalty function method. For C1, the constraint violation value vio1 is calculated as follows:

[0100]

[0101] For C2, the real-number coded individuals need to be discretized. If the value of a dimension is greater than 0.5, the value of the dimension is 1, otherwise it is 0. Then, vio2 is calculated as follows:

[0102]

[0103] For C3, vio3 is calculated as follows:

[0104] vio3=1-dft(G wsn ). (16)

[0105] For C4, vio4 is calculated as follows:

[0106]

[0107] Where Q tj Represents the target t j The specific Q-coverage value of. If C4 is satisfied, then C5 is also satisfied. C6 and C7 can be solved in the encoding stage by controlling the maximum number of target points that each sensor node can cover and the number of other sensor nodes that directly communicate. Combined with the above constraint processing, the fitness value of an individual in stage one is calculated as follows:

[0108]

[0109] Phase 1 divides the deployment area of ​​a large-scale wireless sensor network into multiple independent sub-areas, and each sub-area is optimized by formula (13). Phase 1 guarantees the Q-coverage of the target and the C-connectivity of the sensor nodes in each sub-area. However, all sensor nodes deployed in phase 1 may not satisfy C-connectivity. Therefore, additional sensor nodes need to be deployed in phase 2 to satisfy the C-connectivity of the entire network. The energy consumption of phase 2 is described as follows:

[0110]

[0111] The C1-C4 constraints are consistent with formula (13).

[0112] In summary, the final total energy consumption of the MinEQC problem in large-scale wireless sensor networks is the sum of the energy consumption of stage 1 and stage 2. If all sensor nodes deployed in stage 1 meet C-connectivity, stage 2 does not need to be executed, and the energy consumption of stage 1 is the final total energy consumption.

[0113] 4 Two-stage parallel framework description

[0114] 4.1 Phase 1

[0115] In stage 1, the decision variable in formula (13) represents the deployment strategy of all sensor nodes. Types of sensor nodes deployed in potential deployment locations, then the decision variable The dimension is At the same time, all sensor nodes are deployed in a common The decision variables can be effectively reduced by using the dimensionality reduction method based on neural network. The dimensions are described as follows:

[0116]

[0117] Where X i represents the i-th data in the data set, and W represents the weight of the lightweight network. The dimensionality reduction based on the neural network will The optimization is transformed into the optimization of weight W. For the MinEQC problem, let X i =sn i The features representing all sensor nodes that can be deployed at potential location i constitute a dataset. in represents the coverage radius of sensor nodes of type v, represents the coverage energy consumption of sensor nodes of type v, and represents the coordinates of the sensor node. Deploying different types of sensor nodes at the same potential location may result in similar characteristics. Therefore, let sn i =sn i +N(0,0.01) for sn i A Gaussian perturbation with a mean of zero and a standard deviation of 0.01 is performed. For the neural network in the embedding stage 1, LeakyReLU(x) is used as the activation function of the hidden layer, which is described as follows:

[0118]

[0119] Where α = 0.01. The output layer activation function is described as follows:

[0120]

[0121] MaxFE 1 represents the total number of fitness function evaluations in stage 1, and δ∈[0,1] is used to control the number of fitness function evaluations for reducing the dimension of the neural network in stage 1. Then (1-δ)·MaxFE 1 The number of fitness function evaluations is used by the meta-heuristic algorithm to further optimize the solution based on the solution obtained by the neural network.

[0122] 4.2 Phase 2

[0123] In phase 2, it is necessary to determine whether the independent sub-areas in phase 1 meet C-connectivity. If so, phase 2 does not need to be executed. Otherwise, phase 2 is executed to complete the C-connectivity repair of the large-scale wireless sensor network. In phase 1, by dividing the X and Y axes into and Part, that is, the entire deployment area to be optimized is divided into independent sub-regions. In the second stage, an adaptive connectivity parallel repair strategy based on adjacency sets is proposed. An adjacency set consists of multiple adjacent sub-regions, and different combinations of adjacent sub-regions will produce different numbers of adjacency sets. In order to minimize the number of adjacency sets, if the number of sub-regions allows, four adjacent sub-regions are selected to form an adjacency set; otherwise, two adjacent sub-regions form an adjacency set. The adjacency set is used to determine whether the sub-regions that constitute the adjacency set satisfy C-connectivity. If not, connectivity repair is performed on the adjacency set to ensure the overall C-connectivity of the large-scale wireless sensor network. In this paper, according to different and Value, minimum number of adjacent sets I adjacent Described as follows:

[0124]

[0125] in Represents round down, Z + Represents a non-zero natural number.

[0126] After the adjacency set is determined, a key sensor node is selected from each sub-area that constitutes the adjacency set. These key sensor nodes will be used to form the repair area within the adjacency set. In order to minimize the repair area in the adjacency set, these key sensor nodes should have the shortest Euclidean distance to the geometric center of the adjacency set. Then, the coordinates of the four key nodes are extracted, and the repair area is formed by the maximum and minimum values ​​of each coordinate axis. Finally, it is determined whether the area in each adjacency set needs to be connected and repaired. If not, the adjacency set has satisfied C-connectivity; otherwise, a meta-heuristic algorithm will be used to deploy additional sensor nodes in the repair area according to formula (19). By integrating all the deployed sensor nodes in the first and second stages, the deployment strategy for the entire area can be determined.

[0127] The specific steps of the two-stage parallel framework to solve the MinEQC problem in large-scale wireless sensor networks are as follows:

[0128] Step 1: Get the entire optimization area to be deployed The feature set of all sensor nodes The population size N of the metaheuristic algorithm, the current number of fitness function evaluations FEs, the maximum number of fitness function evaluations MaxFE in stage one and stage two 1 , MaxFE 2 , the proportion δ of the number of evaluations of the neural network fitness function in stage 1, the parameters of the metaheuristic algorithm, and the architecture of the neural network.

[0129] Specific implementation steps for Phase 1:

[0130] Step 2: Set FEs = 0 and set the optimized area to be deployed Divided into multiple independent sub-areas According to the sub-region division method, the feature set Divide

[0131] Step 3: Randomly initialize the weights of the neural network responsible for each sub-region

[0132] %%Parallel start%%

[0133] Step 4: If FEs<δ*MaxFE 1 , go to step 5. Otherwise, go to step 7.

[0134] Step 5: For each independent sub-area According to formula (20), we can obtain a population in stage That is, the deployment strategy of sensor nodes in each sub-area.

[0135] Step 6: Calculate the fitness value of each individual in the population according to formula (18), FEs = FEs + N, and use the meta-heuristic algorithm to Update. Go to step 4.

[0136] Step 7: If FEs≤MaxFE 1 , go to step 8. Otherwise, go to step 9.

[0137] Step 8: Get the current generation According to formula (18), the fitness value of each individual in the population is calculated, FEs = FEs + N, and the meta-heuristic algorithm is used to Update. Go to step 7.

[0138] %%Parallel end%%

[0139] Step 9: Get The optimized area to be deployed Deployment strategy of sensor nodes in phase one.

[0140] Specific implementation steps for Phase 2:

[0141] Step 10: Set FEs = 0, obtain the adjacent set and its number I adjacent , find the key node in each adjacent set, and determine the number of connected recovery areas in the area formed by the key node I restore , I restore ≤I adjacent .

[0142] Step 11: Randomly initialize the population of stage 2

[0143] %%Parallel start%%

[0144] Step 12: If FEs≤MaxFE 2 , go to step 13. Otherwise, go to step 14.

[0145] Step 13: Calculate the fitness value of each individual in the population according to formula (18), FEs = FEs + N, and use the meta-heuristic algorithm to Update. Go to step 12.

[0146] %%Parallel end%%

[0147] Step 14: Integrate all sensor nodes deployed in phases 1 and 2 to form the optimized area to be deployed Final sensor node deployment strategy.

[0148] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, disk, CD-ROM, etc.

[0149] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

[0150] It should be noted that the embodiments of the present invention can be implemented by hardware, software or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will appreciate that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in processor control codes, such as such codes provided on a programmable memory or a data carrier such as an optical or electronic signal carrier.

[0151] In addition, although the operation of the method of the present invention is described in a particular order in the accompanying drawings, this does not require or imply that these operations must be performed in this particular order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps described in the flow chart can change the order of execution. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution. It should also be noted that the features and functions of two or more devices according to the present invention can be embodied in one device. On the contrary, the features and functions of a device described above can be further divided into being embodied by multiple devices.

[0152] Although the invention has been described with reference to several specific embodiments, it should be understood that the invention is not limited to the specific embodiments disclosed. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. An efficient reliability-oriented energy-saving deployment optimization method for ultra-large-scale wireless sensor networks, characterized in that: The method comprises: Obtaining the area to be deployed, dividing the area to be deployed into multiple independent sub-areas and optimizing each sub-area in parallel, and obtaining a feature set of sensor nodes in each sub-area; According to the feature set of sensor nodes in each sub-area, the deployment scheme of each sensor node is obtained by using the forward propagation of the neural network; and the neural network converts the optimization of the sensor node deployment scheme with high-dimensional decision variables into the optimization of the low-dimensional weight matrix of the neural network through dimensionality reduction processing; A meta-heuristic algorithm and a fitness function are used to optimize the low-dimensional weight matrix of the neural network, and the meta-heuristic algorithm obtains a new deployment plan of the sensor nodes according to the optimized low-dimensional weight matrix of the neural network; It is determined whether each sub-region satisfies C-connectivity. When each sub-region does not satisfy C-connectivity, the meta-heuristic algorithm deploys additional sensor nodes to ensure that each sub-region satisfies C-connectivity.

2. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 1 is characterized in that: A meta-heuristic algorithm and a fitness function are used to optimize the low-dimensional weight matrix of a neural network, specifically comprising: the fitness function calculates the deployment plan of each sensor node and obtains a fitness value, and the meta-heuristic algorithm obtains a new low-dimensional weight matrix of the neural network and a deployment plan of each sensor node according to the fitness value.

3. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 2 is characterized in that: When the number of evaluations of the fitness function is less than or equal to the set total number of evaluations, the fitness function is used to calculate the deployment plan of each sensor node and obtain a fitness value, and the meta-heuristic algorithm obtains a new low-dimensional weight matrix of the neural network and a deployment plan of each sensor node according to the fitness value; When the evaluation times of the fitness function is greater than the set total evaluation times, the deployment plans of the sensor nodes in the multiple sub-areas are merged to obtain the deployment plan of the entire area to be deployed.

4. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 3 is characterized by: The dimension reduction process of the neural network also includes Gaussian perturbation and activation function; the Gaussian perturbation is used to prevent the characteristics of different types of sensor nodes at the same potential position from converging; the activation function converts the linear output of neurons in the neural network into nonlinear output by processing the input data of neurons in the hidden layer and output layer.

5. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 4 is characterized in that: The activation function processes the input data of the neurons in the hidden layer and the output layer as follows: For neurons in the hidden layer, the activation function formula is as follows: Where α = 0.01; For the neurons in the output layer, the activation function formula is as follows:

6. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 1 is characterized in that: The specific steps of judging whether C-connectivity is satisfied between each sub-area include: determining an adjacency set consisting of a plurality of adjacent sub-areas, and selecting a key sensor node and the coordinates of the key sensor node from the adjacent sub-areas constituting the adjacency set; forming a repair area according to the coordinates of the key sensor node; and using the judgment result of whether all sensor nodes in the repair area satisfy C-connectivity as the judgment result of whether C-connectivity is satisfied between each sub-area.

7. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 6 is characterized in that: The key sensor node has the shortest Euclidean distance to the geometric center of the adjacency set.

8. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 6 is characterized by: When the number of subregions is sufficient, the adjacent set consists of four adjacent subregions; when the number of subregions is limited, the adjacent set consists of two adjacent subregions.

9. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 7 is characterized in that: The area to be deployed is divided according to the X and Y axes to obtain X×Y sub-areas. According to different X and Y values, the number of adjacent sets I adjacent As shown below: in Represents round down, Z + Represents a non-zero natural number.

10. The efficient reliability-oriented ultra-large-scale wireless sensor network energy-saving deployment optimization method according to claim 1 is characterized in that: The feature set of the sensor node includes the coverage radius of the sensor node, the coverage energy consumption of the sensor node, and the coordinates of the sensor node.

Images