Wireless sensor network coverage optimization method based on improved dung beetle algorithm

Abstract

The invention relates to a wireless sensor network coverage optimization method based on an improved dung beetle algorithm, and belongs to the technical field of wireless sensor network coverage optimization. An existing coverage optimization scheme has a local optimization problem, so that the optimization effect is poor, the coverage blind area is large, and the distribution is not uniform. The improved dung beetle optimization algorithm based on the dung beetle optimization algorithm is realized, and the coverage performance of the algorithm is improved. First, Sobol sequence sampling is used to deploy an initial population to increase diversity. Secondly, a new nonlinear convergence factor is used to balance global search and local mining. And finally, the improved algorithm adopts the Levy flight and crisscross strategy to optimize the later local optimization problem of the algorithm. The method has the advantages of the most uniform network node distribution, the highest network coverage rate and optimization rate and the optimal convergence performance, and has higher adaptability to node optimization deployment in an obstacle environment.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless sensor network coverage optimization, and specifically uses a multi-objective wireless sensor network coverage method based on an improved dung beetle optimization algorithm. Background Art

[0002] WSN is a network composed of multiple micro-sensing nodes. They cooperate with each other to collect, analyze, fuse and transmit data in real time, and can monitor, perceive, collect and process information about the surrounding environment or objects in real time, and perform information interaction through multi-hop communication methods. Wireless sensor network is a research direction with important application value for environment, medical care, intelligent building, agricultural Internet of Things, etc. In practical applications, for different types of nodes in wireless sensor network coverage, a new and challenging method has been proposed.

[0003] In recent years, WSN coverage optimization methods based on swarm intelligence algorithms have received more and more attention. Scholars have proposed the particle swarm algorithm and some other animal algorithms. Both of these algorithms can effectively improve the coverage range and make the node distribution in the network more balanced. However, most of the existing research adopts a single optimization objective, resulting in too high node redundancy, too high node energy consumption, and a short life cycle, and this method itself also has some defects, such as slow convergence speed and easy to fall into local extrema.

[0004] For different service scenarios, when resources such as node energy, communication bandwidth, and hardware are limited, by optimizing the node spatial layout, the network life cycle is extended and the coverage range of the target area is improved. Coverage is an important performance index. On the premise of ensuring the network life cycle, the maximum coverage is achieved by using limited sensing nodes. Therefore, whether the number and layout of sensing nodes in the detection area are reasonable directly affects the QoS of the network. Network coverage optimization studies how to optimize the configuration of network resources through node arrangement and routing selection in a limited network environment. Summary of the Invention

[0005] The objective of the present invention is to, aiming at the disadvantages of traditional swarm intelligence algorithms such as easy to fall into local extrema and poor coverage effect, reduce the energy consumption of useless nodes and extend the network life cycle on the premise of improving the effective coverage range of the network and the node utilization efficiency.

[0006] The technical solution adopted by the present invention is as follows: A wireless sensor network coverage method based on an improved dung beetle optimization algorithm, which uses the improved dung beetle optimization algorithm and Sobol sequence sampling to configure the initial population, thereby improving the diversity of the network. Secondly, a new non-linear convergence coefficient is introduced to enable it to find a balance between global optimization and local optimization. On this basis, a method based on Levy flight and crisscross is proposed to perform subsequent local optimization on this method.

[0007] The specific steps are as follows:

[0008] S1: Construct a node 0 / 1 model in a wireless sensor network. In this model, first, the size of the target area to be monitored by the sensor nodes needs to be determined as A×B, that is, a two-dimensional plane. Next, on this divided two-dimensional plane Area, N identical sensor nodes need to be deployed, and each node has the same radius R to ensure that they can cover the entire area.

[0009] S2: Introduce the Sobol sequence as the initialization means of the algorithm. This diversity is not only reflected in the fact that the population contains various different types of solutions, but also ensures that they have the opportunity to be traversed, thereby improving the global search ability of the algorithm. In this way, the original dung beetle algorithm completes its population initialization.

[0010] S3: When the population initialization is completed, the dung beetles save the best candidate points according to the fitness function, and then update the remaining individuals.

[0011] S4: Since the DBO algorithm has poor ability to jump out of local optima and premature convergence conditions occur during the iteration process. Introducing the Levy flight strategy can significantly improve the global search ability of the algorithm, which is beneficial to escaping local optima, and thus update the candidate solution positions for the next iteration.

[0012] S5: Use a new non-linear factor to balance the exploration and exploitation of the algorithm. In the DBO algorithm, the convergence factor is the balance parameter between the light intensity and the position of the worst dung beetle. When linearly decreasing from 1 to 0.02, the number of dung beetles decreases rapidly. The value of δ drops rapidly to a smaller value, thereby performing a finer search and improving the search accuracy of the algorithm.

[0013] S6: Adopt the vertical and horizontal cross strategy to solve the problem that the population is likely to gather near a local optimum value, which may reduce the global exploration ability of the algorithm. And optimize the DBO algorithm to promote the information exchange between different populations and different dimensions. This strategy improves the ability of the algorithm to escape local optima without affecting its convergence speed.

[0014] S7: Retain the candidate solutions of each individual until the iteration ends at convergence, and return the optimal solution; otherwise, repeat S2 to S7.

[0015] S8: Initialize the population based on the initial coverage positions of N nodes in the WSN, and solve the fitness of the population through the fitness function of this algorithm. On this basis, introduce the improved dung beetle algorithm into the evolutionary process, utilize the horizontal crossover strategy to improve population diversity, accelerate global optimization, and optimize the coverage of the sensing nodes in the detection area.

[0016] Compared with the prior art, the advantages of the present invention include:

[0017] The present invention optimizes the deployment of WSN sensor nodes based on the improved dung beetle algorithm, deploys the sensor nodes in the monitoring area, and tests the effectiveness of the optimization of the improved dung beetle algorithm by changing the number of nodes, the sensing radius of the nodes, and the area of the monitoring area. The results show that the improved dung beetle algorithm improves the coverage performance of the wireless sensor network. It achieves higher coverage and optimization rates with fewer nodes or a smaller sensing radius, reducing the deployment cost. The node deployment optimized by the improved dung beetle algorithm is more uniform, resulting in the minimum positioning error of the algorithm.

[0018] Secondly, the improved dung beetle algorithm introduces Sobol sequence sampling to initialize the population, enhancing the diversity within the population. Subsequently, it adopts the Levy flight strategy and vertical and horizontal crossover strategies to solve the late local optimization problem in the dung beetle algorithm. Finally, a non-linear convergence factor balances the global and local search capabilities of the algorithm. Through simulation experiments, the improved dung beetle algorithm proves to have better convergence speed, accuracy, and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is the flowchart of the wireless sensor network coverage method of the improved dung beetle optimization algorithm of the present invention;

[0020] Figure 2 are the convergence curves and box plots of the experiments of 5 optimization algorithms;

[0021] Figure 3 are the node coverage result diagrams of the five algorithms of PSO, SCA, SSA, DBO, and IDBO; Figure 3 Among them: (a) is the initialized node coverage diagram; (b) is the node coverage diagram of the SCA algorithm; (c) is the node coverage diagram of the PSO algorithm; (d) is the node coverage diagram of the SSA algorithm; (e) is the node coverage diagram of the DBO algorithm; (f) is the node coverage diagram of the IDBO algorithm. DETAILED DESCRIPTION OF THE INVENTION

[0022] The present invention will be further described below in conjunction with the drawings and specific embodiments.

[0023] Specifically, S1 is as follows:

[0024] S101: Set the monitoring area as a rectangle with an area of A×B, and n homogeneous sensor nodes are randomly distributed, defined as S = {S 1 , S 2 , L, S i , L, S n}}. Each node has the same radius R S , where S i = (x i , y i ), i = 1, 2, ……, n. After discretizing the monitoring area, the coordinate position of each sub-area is C j = (x j , y j ), j = 1, 2, L, p×q. Then, the Euclidean distance from the sensor node to the target point can be expressed as:

[0025]

[0026] Adopt the probability perception model to more realistically reflect the actual perception effect of the sensor node on the target:

[0027]

[0028] where P(S i , C j ) is the perception probability, R u represents the perception range where the perception probability decays exponentially, α = R u - R s + d(S i , C j ), indicating the degree to which the perception probability decreases with the increase of distance, and λ and β represent the measurement parameters connected to the node. If multiple sensor nodes can sense the target point, the joint perception probability is defined as follows:

[0029]

[0030] S102: P cov is the area coverage rate. The WSN coverage rate can be defined as the ratio of the number of targets covered by the sensor set to the total number of targets in the area as follows:

[0031]

[0032] S2: The position update formula of the rolling ball dung beetle behavior is shown in the figure:

[0033]

[0034] is the position of the nth dung beetle at the tth iteration, where t represents the number of iterations, η is a natural number 0 or 1, 1 means the movement of the dung beetle does not deviate from the original direction, -1 means the movement of the dung beetle completely deviates from the original direction, and ξ is a constant in the range (0, 0.2], is used to simulate the change of light intensity, represents the worst position of the current population, and δ is a constant. The calculation formula is as follows:

[0035]

[0036] Maxiter is the maximum number of iterations.

[0037] Use the Sobol sequence to create a random number K n ∈[0, 1]. Assume that the optimal solution has a value range [x min , x max . Then the initial position of the improved dung beetle can be characterized as follows:

[0038] x new = x min + K n ·(x max - x min ) (7)

[0039] S3: When the population initialization is completed, the dung beetles save the best candidate points according to the fitness function f(x new ), and then update the remaining individuals. The WSN coverage problem is transformed into finding the optimal solution of P cov as described below:

[0040] f(P) = max[P cov (P)] (8)

[0041] S4: Using the Levy flight strategy can significantly improve the global search ability of the algorithm and is beneficial to escaping from local optima. If the difference is less than the given threshold, it means that the algorithm has fallen into a local optimum, and then the Levy flight strategy is adopted:

[0042]

[0043] where d is the dimension, L is the search step size, is the number solution of the ith time at the tth iteration, is the global best solution of the entire population. The calculation method of the levy function is as follows:

[0044]

[0045] where λ 1 and λ 2is a random number on [0, 1], β is a constant, and in this paper, 1.5 is taken. σ is calculated as follows:

[0046]

[0047] where Γ(x) = (x - 1)! and x belongs to natural numbers.

[0048] S5: The convergence factor δ is a balance parameter between the light intensity and the worst dung beetle position. When δ linearly decreases from 1 to 0.02, the number of dung beetles rapidly decreases, which will lead to too high local population density, reduced diversity, premature convergence, and quickly falling into the local optimum. A new non - linear factor is proposed to balance the exploration and exploitation of the algorithm, and its mathematical description is as follows:

[0049]

[0050] S6: The algorithm must balance the local search and global exploration capabilities to avoid falling into the local optimum state and prevent premature convergence. A longitudinal and transverse crossover strategy is adopted to solve this problem, and this strategy improves the ability of the algorithm to escape the local optimum without affecting its convergence speed. Horizontal crossover: Initially, the parent individuals are randomly paired in dimensions, and the crossover operation is as follows:

[0051]

[0052] where r 1 and r 2 are random values uniformly distributed between 0 and 1; c 1 and c 2 are random values uniformly distributed between - 1 and 1; and are respectively and representations of the offspring.

[0053] Vertical crossover: The vertical crossover operation occurs less frequently than the horizontal crossover operation and affects all dimensions of the offspring. It is similar to the mutation operation in the genetic algorithm:

[0054]

[0055] where r is a random value uniformly distributed between 0 and 1, is and offspring.

[0056] S7: Calculate f(P) to retain the candidate solutions of each individual, judge whether the number of iterations reaches the maximum number of iterations, end the iteration until convergence, and return the optimal solution, otherwise repeat Step2 to Step7.

[0057] The present invention will be further elaborated through specific examples below.

[0058] Simulation experiment and result analysis: To verify the application effect of the WSN coverage optimization based on the improved dung beetle algorithm, the algorithm of the present invention is selected for comparative experiments with the basic particle swarm optimization algorithm (PSO), the sparrow search algorithm (SSA), the basic dung beetle optimization algorithm (DBO), and the sine-cosine algorithm (SCA). The experimental area is set to 10,000 m 2 , the number of nodes is 150, the sensor radius R = 5 m. The improved algorithm proposed by the present invention can reduce node overlap, the coverage rate reaches 98.21%, which is better than other algorithms, and the convergence speed is also faster.

Claims

1. A wireless sensor network coverage optimization method based on an improved dung beetle swarm algorithm is characterized by: The following steps are involved: S1: Construct a node 0 / 1 model in a wireless sensor network. In this model, we first need to establish that the target area to be monitored by the sensor node is A×B, that is, a two-dimensional plane. Next, N identical sensor nodes need to be deployed on this divided two-dimensional plane Area, and each node has the same radius R to ensure that they can cover the entire area. S2: Introduce the Sobol sequence as a means of initializing the algorithm. This diversity is not only reflected in the inclusion of various types of solutions in the population, but also ensures that they have the opportunity to be traversed, thereby improving the global search ability of the algorithm. In this way, the original dung beetle algorithm is able to complete its population initialization. S3: When the group initialization is completed, the dung beetle saves the best candidate points according to the fitness function, and then updates the remaining individuals. S4: Due to the poor ability of the DBO algorithm to escape from the local optimum, premature convergence conditions occurred during the iteration process. The introduction of the Levy flight strategy can significantly improve the global search ability of the algorithm, which is conducive to escaping the local optimum and updating the candidate solution position for the next iteration. S5: Use a new nonlinear factor to balance the exploration and exploitation of the algorithm. In the DBO algorithm, the convergence factor is a balance parameter between light intensity and the worst dung beetle position. When it is linearly reduced from 1 to 0.02, the number of dung beetles decreases rapidly. The value of δ drops rapidly to a smaller value, resulting in a more refined search and improving the search accuracy of the algorithm. S6: A vertical and horizontal crossover strategy is used to solve the problem that the population is likely to gather near a local optimal value, which may reduce the global exploration ability of the algorithm. The DBO algorithm is optimized to promote information exchange between different populations and different dimensions. This strategy improves the algorithm's ability to escape the local optimal without affecting its convergence speed. S7: Keep the candidate solution of each individual until the iteration ends when convergence occurs and return to the optimal solution, otherwise repeat S2 to S7. S8: The group is initialized based on the initial coverage positions of N nodes in WSN, and the fitness of the group is solved through the fitness function of the algorithm. On this basis, the improved dung beetle algorithm is introduced into the evolution process, and the horizontal crossover strategy is used to improve the diversity of the group, accelerate the global optimization, and optimize the coverage of the sensing nodes in the detection area.

2. The wireless sensor network coverage optimization method based on the improved dung beetle swarm algorithm according to claim 1 is characterized in that: The S1 is specifically: S101: Set the monitoring area to be a rectangle with an area of ​​A×B, and n homogeneous sensor nodes are randomly distributed, defined as S={S1,S2,L,S i ,L,S n }. Each node has the same radius R S , where S i =(x i ,y i ), i = 1, 2, ..., n. After the monitoring area is discretized, the coordinate position of each sub-area is C j =(x j ,y j ), j = 1, 2, L, p × q, then, the Euclidean distance from the sensor node to the target point can be expressed as: The probabilistic perception model is used to more realistically reflect the actual perception effect of the sensor node on the target: Where P(S i ,C j ) is the perception probability, R u The perception range α = R represents the exponential decay of the perception probability u -R s +d(S i ,C j ), represents the degree to which the perception probability decreases with increasing distance, and λ and β represent the measurement parameters connected to the node. If multiple sensor nodes can sense the target point, the joint perception probability is defined as follows: S102:P cov is the area coverage rate. The WSN coverage rate can be defined as the ratio of the number of targets covered by the sensor set to the total number of targets in the area as follows:

3. The wireless sensor network coverage optimization method based on the improved dung beetle swarm algorithm according to claim 1 is characterized in that: The S2 is specifically: S201: The position update formula of the rolling dung beetle behavior is shown in the figure: X n t is the position of the dung beetle at the nth iteration, t represents the number of iterations, η is a natural number 0 or 1, 1 means that the movement of the dung beetle does not deviate from the original direction, -1 means that the movement of the dung beetle completely deviates from the original direction, ξ is a constant in the range of (0,0.2], is used to simulate changes in light intensity. Represents the worst position of the current population, δ is a constant, and the calculation formula is as follows: Maxiter is the maximum number of iterations. S202: Create a random number K using the Sobol sequence n ∈[0,1], assuming that the optimal solution has a value range [x min , x max ]Then the initial position of the improved dung beetle can be characterized as follows: x new =x min +K n ·(x max -x min ) (7) 4. The wireless sensor network coverage optimization method based on the improved dung beetle swarm algorithm according to claim 1 is characterized in that: The S3 is specifically: S3: When the population is initialized, the dung beetles follow the fitness function f(x new ), save the best candidate points, and then update the remaining individuals. The WSN coverage problem is transformed into finding P cov The optimal solution is described as follows: f(P)=max[P cov (P)] (8) 5. The wireless sensor network coverage optimization method based on the improved dung beetle swarm algorithm according to claim 1 is characterized in that: The S4 is specifically: S4: Using the Levy flight strategy can significantly improve the global search capability of the algorithm and help escape the local optimum. If the difference is less than the given threshold, it means that the algorithm has fallen into the local optimum, and the Levy flight strategy is adopted: In the formula, d is the dimension, L is the search step length, is the ith scalar solution at the tth iteration, is the global optimal solution for the entire population. The calculation method of the levy function is as follows: Where λ1 and λ2 are random numbers on [0,1], β is a constant, which is 1.5 in this paper, and σ is calculated as follows: Among them, Γ(x)=(x-1)! And x is a natural number.

6. The wireless sensor network coverage optimization method based on the improved dung beetle swarm algorithm according to claim 1 is characterized in that: The S5 is specifically: S5: The convergence factor δ is a balance parameter between light intensity and the worst dung beetle position. When δ decreases linearly from 1 to 0.02, the number of dung beetles decreases rapidly, resulting in excessive local population density, reduced diversity, premature convergence, and rapid decline into the local optimum. A new nonlinear factor is proposed to balance the exploration and exploitation of the algorithm, and its mathematical description is as follows:

7. The wireless sensor network coverage optimization method based on the improved dung beetle swarm algorithm according to claim 1 is characterized in that: The S6 is specifically: S601: The algorithm must balance local search and global exploration capabilities to avoid falling into local optimal states and prevent premature convergence. A vertical and horizontal crossover strategy is used to solve this problem, which improves the algorithm's ability to escape local optimality without affecting its convergence speed. Horizontal crossover: Initially, the parent individuals are randomly paired in the dimension and the crossover operation is performed as follows: Where r1 and r2 are random values ​​uniformly distributed between 0 and 1; c1 and c2 are random values ​​uniformly distributed between -1 and 1; and They are and Representation of offspring. S602: Vertical crossover: The vertical crossover operation occurs less frequently than the horizontal crossover operation and affects all dimensions of the offspring. It is similar to the mutation operation in the genetic algorithm: Among them, r is a random value uniformly distributed between 0 and 1. for and Offspring of. S7: Calculate f(P) and retain the candidate solution of each individual, determine whether the number of iterations reaches the maximum number of iterations, end the iteration until convergence, and return the optimal solution, otherwise repeat S2 to S7.

Images