CN108810914A - Based on the WSN Node distribution optimization methods for improving weeds algorithm - Google Patents

Abstract

The invention discloses a WSN node distribution optimization method based on the improved weed algorithm. The method solves the optimal distribution of WSN nodes based on the improved weed algorithm, and the improved weed algorithm introduces a cubic mapping chaos operator to improve the local search ability of the algorithm. Using the Gaussian mutation operator to enhance the diversity of the population has the advantages of fast convergence, good robustness, and strong data mining capabilities, and can effectively solve the problem of WSN node distribution optimization.

Description

WSN Node Distribution Optimization Method Based on Improved Weed Algorithm

technical field

The invention relates to a WSN node distribution optimization method, in particular to a WSN node distribution optimization method based on an improved weed algorithm.

Background technique

Wireless sensor networks (WSN for short) is a multi-hop wireless network composed of a group of stationary or mobile sensor nodes in the form of self-organization, which has the advantages of strong invulnerability and rapid deployment. It has a wide range of application prospects, and has been a hot research topic at home and abroad in recent years. The research shows that the reasonable arrangement of sensor nodes is beneficial to improve the comprehensive performance of WSN, but it is also prone to channel interference and information redundancy, resulting in energy waste. Therefore, research on how to reasonably deploy sensor nodes and optimize network performance has become one of the key technologies of WSN.

For the optimization of WSN node distribution, many scholars at home and abroad use artificial intelligence algorithms to deal with it. Common AI algorithms include: Genetic algorithm (GA for short), Particle swarm optimization algorithm (PSO for short), Artificial bee colony algorithm (ABC for short), etc. However, almost all intelligent algorithms are prone to fall into local optimum, and there are problems of premature convergence in the early stage and slow convergence speed in the later stage.

Contents of the invention

In order to solve the above technical problems, the present invention provides a WSN node distribution optimization method based on the improved weed algorithm.

In order to achieve the above object, the technical scheme adopted in the present invention is:

The WSN node distribution optimization method based on the improved weed algorithm includes the following steps,

1) Construct the WSN node distribution optimization objective function;

2) Solve the optimal distribution of WSN nodes based on the improved weed algorithm;

The specific process is as follows:

21) Population initialization: use the cubic map chaos operator to randomly generate weed positions;

22) Growth and reproduction: calculate the fitness value of each weed, and calculate the number of seeds produced by each weed;

23) Spatial diffusion: The seeds are randomly scattered in the neighborhood of their parent weeds with a normal distribution, and some of the seeds use the Gaussian mutation operator to generate new mutant weeds;

24) Competitive elimination: After several iterations, when the population exceeds the maximum population size P _max , all weeds are sorted according to their fitness values from large to small, and the top P _max weeds are retained;

25) Stop criterion: Repeat steps 22 to 24, record the individual with the best fitness value in each generation population, until the number of iterations reaches the maximum number of iterations, the algorithm stops, and the optimal solution of iteration is output, which is the optimal distribution of WSN nodes.

WSN node distribution optimization objective function C,

Among them, the definition area A is discretized into s×s grids, n is the number of WSN nodes in the area, m×n is the number of grids, C _j is the probability that grid point p _j is detected by WSN nodes, C _ij is the measurement model of WSN node s _i ,

Among them, d _ij is the distance between grid point p _j and WSN node s _i , λ ₁ , λ ₂ , ε ₁ , ε ₂ are measurement parameters, α ₁ =R _e -R+d _ij , α ₂ =R _e +Rd _ij , _Re is the effective measurement radius of the WSN node, and R is the sensing radius of the WSN node.

Using the cubic map chaos operator to randomly generate the weed position formula is,

X _i′ =[ _xi′1 , _xi′2 ,…, _xi′D ]

Among them, X _i′ is the position of weed i′, D is the dimension of weed i′, D∈[1,D], x _i′d is the position of weed i′ in dimension d, x _U and x _L are the upper and lower limits of x _i′d values, and y _i′d is the d-th dimension value of weed i′ generated by chaotic sequence.

The formula for the number of seeds produced by each weed is,

Among them, P _s is the number of seeds, f _max , f _min are the maximum fitness value and minimum fitness value of the population, S _max , S _min are the maximum and minimum seed numbers, f(X _i′ ) is the weed i′ fitness value.

The position V _i′ of the mutated weed is,

V _i' =X _i' +e(X _B -X _i' )

Among them, e is a Gaussian distribution with a mean of 0 and a variance of 1, and Xi _' is the position of the parent weed of the mutant weed, that is, the position of weed i'.

When the position V _i' of the mutated weed is not within the search range, the position V _i' of the mutated weed is,

V _i′ ＝X _L +e(X _U -X _L )

Among them, X _U and X _L are the upper and lower limits of the value of Xi _' respectively.

The beneficial effect achieved by the present invention: the present invention is based on the improved weed algorithm to solve the optimal distribution of WSN nodes, the improved weed algorithm introduces the cubic mapping chaos operator to improve the local search ability of the algorithm, and uses the Gaussian mutation operator to enhance the diversity of the population It has the advantages of fast convergence speed, good robustness, and strong data mining ability, and can effectively solve the problem of WSN node distribution optimization.

Description of drawings

Fig. 1 is a flowchart of the present invention;

Figure 2 is the iterative curve of the optimal solution of the network coverage of the three algorithms;

Figure 3 is a comparison chart of network coverage;

Fig. 4 is the initial WSN node distribution diagram;

Figure 5 is the optimized WSN node distribution diagram.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solution of the present invention more clearly, but not to limit the protection scope of the present invention.

As shown in Figure 1, the WSN node distribution optimization method based on the improved weed algorithm includes the following steps:

1) Construct WSN node distribution optimization objective function.

Define the two-dimensional plane area A to be discretized into s×s grids, and the area of each grid is 1. If there are n WSN nodes (that is, wireless sensor nodes) distributed in the whole area, each WSN node can pass some kind of A special way (like GPS) to get its own position, and have the same perception radius R. Therefore, the set of all WSN nodes in area A can be described as:

S＝{s ₁ ,s ₂ ,…,s _n }

Wherein, the coordinates of WSN node s _i are ( _xi ,y _i ), i∈[1,n].

The distance between grid point p _j and WSN node s _i is,

Among them, j∈[1,m×n], m×n is the number of grids, and the coordinates of grid point p _j are (x _j ,y _j ).

There are two main measurement models for WSN node s _i , one is a binary measurement model, and the other is a probability measurement model. Here, the probability measurement model is used.

Among them, C _ij is the measurement model of WSN node s _i , λ ₁ , λ ₂ , ε ₁ , and ε ₂ are measurement parameters, α ₁ =R _e -R+d _ij , α ₂ =R _e +Rd _ij .

If all WSN nodes are detected as independent events, then the probability of grid point p _j being detected by WSN nodes is:

Among them, if C _j is greater than or equal to a certain threshold C _t , it is considered that the grid point p _j can be detected by the node; otherwise, if C _j is less than C _t , it is considered that the grid point p _j cannot be detected, and this paper chooses _Ct = 0.75.

The coverage rate of each grid is measured by the probability that the grid point p _j is detected, and the ratio of the number of detected grid points to the total number of grids is taken as the coverage rate C _s of the WSN. The specific formula is,

Therefore, the WSN node distribution optimization problem can be described as n WSN nodes in area A achieve coverage of the entire target area through an optimization algorithm. That is, this problem can be transformed into a coverage maximization problem, that is, the objective function of WSN node distribution optimization is,

Among them, C is the objective function of WSN node distribution optimization.

2) Solve the optimal distribution of WSN nodes based on the improved weed algorithm (IIWO algorithm).

The specific process is as follows:

21) Population initialization: use the cubic map chaos operator to randomly generate weed positions.

In the weed algorithm (IWO algorithm), randomly generating the initial position of the weed may lead to uneven distribution of the position. Considering that the chaotic operator has the characteristics of randomness and regularity, and can not repeatedly traverse all the weeds within a certain range State, in the chaotic model, the sequence generated by the cubic map is more uniform than the commonly used Logistic map, so the cubic map chaotic operator is used here to improve the initialization of the weed position.

The specific process of initialization is as follows:

a) For M individuals in the D-dimensional space, randomly generate a D-dimensional vector Y;

b) Use the chaotic sequence to perform M-1 iterations on Y dimension by dimension, which produces the remaining M-1 individuals;

y(k+1)=4y(k) ³ -3y(k)(k=0,1,…)

Among them, k is the number of iterations, y(k) is the chaotic sequence at the kth iteration, and y(k+1) is the chaotic sequence at the k+1th iteration;

c) Map the generated chaotic variables into the search space of the solution according to the following formula:

Among them, D∈[1,D], x _i′d is the position of weed i′ in dimension d, x _U and x _L are the upper and lower limits of x _i′d respectively, and y _i′d is the The d-th dimension value of the weed i′ produced.

The weed position formula obtained by initialization is,

X _i′ =[ _xi′1 , _xi′2 ,…, _xi′D ]

Among them, Xi _' is the position of weed i', and D is the dimension of weed i'.

22) Growth and reproduction: calculate the fitness value of each weed, and calculate the number of seeds produced by each weed; the better the population, the more seeds the weed produces,

The formula for the number of seeds produced by each weed is,

Among them, P _s is the number of seeds, f _max , f _min are the maximum fitness value and minimum fitness value of the population, S _max , S _min are the maximum and minimum seed numbers, f(X _i′ ) is the weed i′ fitness value.

23) Spatial Diffusion: Seed in Normal The distribution is randomly scattered in the neighborhood of its parent weeds, and some of the seeds use the Gaussian mutation operator to produce new mutant weeds.

During the iterative process of the algorithm, the change rule of the standard deviation can be described as:

Among them, w is the nonlinear adjustment factor, σ _I and σ _F are the initial standard deviation and final standard deviation of sowing seeds respectively, σ _iter is the standard deviation of the iter generation, iter is the number of iterations, and iter _max is the maximum number of iterations.

In order to avoid the IWO algorithm from falling into the local optimum, some seeds are randomly selected, and Gaussian mutation operators are used to generate new mutant weeds. The position V _i′ of the mutant weeds is,

V _i' =X _i' +e(X _B -X _i' )

Among them, e is a Gaussian distribution with a mean of 0 and a variance of 1, Xi _' is the position of the parent weed of the mutant weed, that is, the position of the weed i', and X _B is the weed with the highest fitness value in the current population Location.

There is a Gaussian random interference item between the parent generation and the current optimal weed in the mutant weed, which may cause the location of the mutant weed to exceed the search range of the algorithm. Therefore, when the position V _i' of the mutated weed is not within the search range (that is, the search range of the IWO algorithm), the position V _i' of the mutated weed is,

V _i′ ＝X _L +e(X _U -X _L )

Among them, X _U and X _L are the upper and lower limits of the value of Xi _' respectively.

24) Competitive elimination: After several iterations, when the population exceeds the maximum population size P _max , all weeds are sorted according to the fitness value from large to small, and the top P _max weeds are retained.

25) Stop criterion: Repeat steps 22 to 24, record the individual with the best fitness value in each generation population, until the number of iterations reaches the maximum number of iterations, the algorithm stops, and the optimal solution of iteration is output, which is the optimal distribution of WSN nodes.

In order to test the performance of the above-mentioned improved weed algorithm (IIWO algorithm), the literature (Lei L, Shiru Q. Path Planning for Unmanned Air Vehicles Using an Improved Artificial Bee Colony Algorithm[C]//Control Conference (CCC), 2012 31st Chinese.IEEE , 2012:2486-2491.) to test the four standard test functions, and compare with the CPSO algorithm (Chaotic PSO, CPSO for short) and the IWO algorithm. The theoretical global minimum values of the four functions are all 0, and the simulation environment is Windows 7 operating system and MATLAB R2016a compiled software. In order to compare the optimization performance of each algorithm fairly, the population size of each algorithm is set to 40, the dimension of each test function is 30, and the number of iterations is 500. Then, 30 independent tests are carried out on each test function by each algorithm. Table 1 shows the statistical results of the test, including the mean (Mean) and standard deviation (SD).

Table 1 Standard function test results

It can be seen from the table that the CGSO algorithm is superior to the IWO algorithm and the CPSO algorithm in terms of the Sphere, Ackley and Rastrigin functions, both in terms of average and standard deviation, and has a significant order of magnitude improvement. Compared with IWO algorithm and CPSO algorithm, IIWO algorithm improves the Sphere function by 40 and 2 orders of magnitude respectively. Although the performance of the IIWO algorithm on the Griewank function is slightly lower than that of the CPSO algorithm, the mean and standard deviation of the two are still in the same order of magnitude. This shows that compared with the other two algorithms, the IIWO algorithm can fully develop the information of the searched object and has a higher solution quality.

In order to test the effectiveness of the IIWO algorithm in dealing with the WSN node distribution optimization problem, this simulation sets the effective monitoring range of the WSN to a square area of 100m×100m, the sensing radius of each sensor node is 7m, and 100 sensor nodes are randomly distributed within the monitoring range. WSN nodes. The layout of these nodes is optimized by IWO algorithm, CPSO algorithm and IIWO algorithm respectively, iterating 500 generations, repeating the operation 30 times, and recording the best solution among them. In addition, the population numbers of the three algorithms are all set to 20, and the settings of other parameters are shown in Table 2.

Table 2 Parameter settings of the three algorithms

After the algorithm runs, the optimal coverage rate and average coverage rate of each algorithm are counted, and the results are shown in Table 3. It can be seen from the table that the IWO algorithm improved by the chaos operator and the Gaussian mutation operator has greatly improved the success probability of WSN distribution optimization. The average WSN coverage obtained by IIWO algorithm is 2.8% and 13.3% higher than the other two algorithms respectively.

Table 3 Comparison of the results of the three algorithms

The iterative curves of the three algorithms to deal with the optimal solution of network node coverage are shown in Figure 2. It can be seen from the figure that the IWO algorithm and the CPSO algorithm began to converge at 479 and 197 iterations respectively, while the IIWO algorithm began to converge at 95 iterations. In addition, the optimal network coverage obtained by the algorithm in this paper is 99.39%, while those obtained by the CPSO algorithm and the IWO algorithm are 99.08% and 97.75%, respectively. This shows that the IIWO algorithm has fast convergence speed, strong global search ability, and can get rid of the constraints of local optimum.

In addition, in order to verify the impact of the population size on the performance of the algorithm, the population size is set to 10, 20 and 30, and each iteration is 300 generations, and the next algorithm is finally recorded to obtain the network coverage, as shown in Table 4. It can be seen from the table that with the expansion of the population size, the optimization accuracy of the IIWO algorithm is higher than that of other algorithms. This shows that the algorithm proposed in this paper can obtain more search information and maintain the diversity of the population, and has strong robustness.

Table 4 Comparison of optimization results of each algorithm under different population sizes

In order to further verify the performance of the IIWO algorithm, a network coverage simulation based on different numbers of nodes is designed. The change curves of the network coverage rate of the three algorithms with the node density are shown in Figure 3. It can be seen from the figure that to achieve a coverage rate of more than 95%, the CPSO algorithm and the IWO algorithm need to deploy 150 and 200 nodes respectively, while the IIWO algorithm only needs to deploy 125 nodes. This shows that compared with the other two algorithms, the IIWO algorithm has a stronger ability to mine global information.

It can be seen from Figure 4 and Figure 5 that after the nodes in the initial state are disordered and arranged optimally by the IIWO algorithm, the distribution is relatively uniform, and the overlapping coverage is relatively small. Therefore, the WSN node distribution optimization method based on the improved weed algorithm proposed in this paper can reasonably solve the network coverage optimization problem, and can effectively improve the network coverage.

The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, some improvements and modifications can also be made. It should also be regarded as the protection scope of the present invention.

Claims (6)

1. based on the WSN Node distribution optimization methods for improving weeds algorithm, it is characterised in that：Include the following steps,

1) WSN Node distribution optimization object functions are built；

2) it is based on improving weeds algorithm solution WSN node Optimal Distributions；

Detailed process is as follows：

21) initialization of population：Weeds position is randomly generated using cube mapping chaos operator；

22) growth and breeding：The fitness value for calculating each weeds calculates the seed number that each weeds generate；

23) space is spread：Seed is randomly dispersed with normal distribution in the neighborhood of its parent weeds, and which part seed utilizes height This mutation operator generates new variation weeds；

24) competition is eliminated：After iteration several times, when population number is more than maximum population scale number P_maxWhen, all weeds according to Fitness value is ranked up from big to small, P before retaining_maxA weeds；

25) stopping criterion：Step 22~24 are repeated, the individual that fitness value is best in per generation population is recorded, until iteration time Number reaches greatest iteration number, and algorithm stops, and exports iteration optimal solution, as WSN nodes Optimal Distribution.

2. according to claim 1 based on the WSN Node distribution optimization methods for improving weeds algorithm, it is characterised in that：WSN Node distribution optimization object function C,

Wherein, definition region A is separated into the grid of s × s, and n is WSN node numbers in region, and m × n is the number of grid, C_jFor Mesh point p_jThe probability arrived by WSN nodal tests,C_ijFor WSN nodes s_iMeasurement model,

Wherein, d_ijFor mesh point p_jWith WSN nodes s_iDistance, λ₁,λ₂,ε₁,ε₂It is measurement parameter, α₁=R_e-R+d_ij, α₂= R_e+R-d_ij, R_eFor effective measurement radius of WSN nodes, R is the perception radius of WSN nodes.

3. according to claim 1 based on the WSN Node distribution optimization methods for improving weeds algorithm, it is characterised in that：Profit Randomly generating weeds location formula with cube mapping chaos operator is,

X_i′=[x_i′1,x_i′2,…,x_i′D]

Wherein, X_i′For the position of weeds i ', D is the dimension of weeds i ', D ∈ [1, D], x_i′dIt is weeds i ' in the position that d is tieed up, x_U And x_LRespectively x_i′dThe bound of value, y_i′dD dimension values for the weeds i ' generated using chaos sequence.

4. according to claim 1 based on the WSN Node distribution optimization methods for improving weeds algorithm, it is characterised in that：Often The seed number formula that a weeds generate is,

Wherein, P_sFor seed number, f_max,f_minRespectively population maximum adaptation angle value and minimum fitness value, S_max,S_minRespectively Minimum and maximum seed number, f (X_i′) be weeds i ' fitness value.

5. according to claim 1 based on the WSN Node distribution optimization methods for improving weeds algorithm, it is characterised in that：Become The position V of different weeds_i′For,

V_i′=X_i′+e(X_B-X_i′)

Wherein, it is 0 that e, which is mean value, the Gaussian Profile that variance is 1, X_i′For the parent weeds position for the weeds that make a variation, the i.e. position of weeds i ' It sets.

6. according to claim 5 based on the WSN Node distribution optimization methods for improving weeds algorithm, it is characterised in that：When The position V of variation weeds_i′When not in search range, then the position V for the weeds that make a variation_i′For V_i′=X_L+e(X_U-X_L)

Wherein, X_UAnd X_LRespectively X_i′The bound of value.