CN110750754A - Data processing method based on wireless sensor network in big data environment - Google Patents

Abstract

A data processing method based on a wireless sensor network in a big data environment is provided. The existing wireless sensor network has the problems of poor accuracy and real-time performance of obtained data, node energy consumption and data post-processing capability. The invention innovatively utilizes the research of a wireless sensor network system to process the delay and loss conditions of data in the system, thereby improving the data transmission precision; a variable definition method is innovatively adopted, and the system is subjected to the amplification and dimension reduction treatment on the basis of a linear matrix inequality method, so that the method is easy to operate and improve the operation speed; the estimation algorithm proposed is more accurate than Kalman filtering.

Description

Data processing method based on wireless sensor network in big data environment

Technical Field

The invention relates to a data processing method in a big data environment based on a wireless sensor network.

Background

Big data is large-scale, particularly complex data, and statistically, the data generated in recent years accounts for 90% of the total data generated by human beings. Meanwhile, global sensing equipment collects information in real time, the data sources are countless, and various sensors such as the internet of things, the mobile internet, cloud computing, personal computers and mobile phones form a complex big data background. A great deal of research is conducted internationally and domestically, the united states uses big data technology to control world data, korea draws up a comprehensive growth plan for cloud computing, amazon introduced desktops and services, IBM introduced private cloud services, and so on. A cloud computing data platform, a Beijing cloud base, a Hangzhou cloud computing service platform, a tin-free cloud computing center and the like are built in part of provinces and cities in China. However, with the increasing amount of data, the traditional mining and storage forms are far from meeting the demand.

The research of Wireless Sensor Networks (WSNs) has just started in the late 90 s of the 20 th century, and has attracted great attention in military, academic and industrial fields in the early 21 st century. The 'national intelligent transportation system project planning' is proposed by the U.S. department of transportation in 1995, and attempts are made to integrate technologies such as information, computers, sensors and the like and preliminarily apply the integrated technologies to a ground transportation management system, so that high efficiency, omnibearing and integrated technologies are realized. The concept of wireless sensor network is firstly proposed in the united states of the seventies of the twentieth century, and then some military and civil wireless sensor network monitoring systems are developed. Since the twenty-first century, the rapid development of sensor technology, microelectronic technology, wireless communication technology, integrated circuit technology, computer technology, and the like has provided a good foundation for the development of wireless sensor networks, and wireless sensor network technology with low power consumption, low cost, and miniaturization has gained a chance of rapid development. Since the research enthusiasm has been raised worldwide, various organizations have fully realized that the wireless sensor network has huge application prospect and commercial value, and governments begin to invest a great deal of expenses for the related research of the wireless sensor network.

In the end of the twentieth century, colleges and scientific research institutions in China begin to explore and research wireless sensor networks, and countries begin to provide corresponding fund funding. In 2006, China releases a compendium for the development of the long-term science and technology in China, emphasizes the acceleration of the research and development of information technology in the compendium for the planning, and particularly proposes the acceleration of the research and application of a wireless sensor network. The intelligent perception and self-organizing network technology related to the technology is listed as the leading-edge technology of the information technology, and a plurality of national scientific research institutions are used as key points to research and develop. Although research on the technical field of WSNs starts somewhat later in China than abroad, research efforts have been increased on WSNs by countries and research institutes. Research is being conducted in the field of WSNs by some scientific research institutions and general universities. The national 863 program and the like also actively pursue and support research on various technologies of WSNs. The WSN is a brand-new information acquisition and processing technology applied to a plurality of fields, the gap between the domestic and foreign aspects of the research level and the work of the WSN is not large, the technical research of the WSN is timely and reasonably developed, the future study, life, work and the like of people are greatly influenced, and the national development, social progress and economic prosperity can be promoted.

The application technology of big data is combined with a wireless sensor network, and is already in logistics, retail sale, communication, medical treatment and electric power

And the like, are widely used and play an important role. Because the wireless sensor is the core part of the Internet of things, a large number of sensor nodes continuously transmit acquired data to the data center to form a mass data stream. The processing and storage of mass data is a great problem to be faced by the wireless sensor network, but currently, related technical research on the management and processing of mass data of the application layer of the wireless sensor network is still relatively few, and there is a great research and development space. In a wireless sensor network, the performance of a data sensing/acquisition technology directly affects the accuracy and real-time performance of acquired data, node energy consumption and data post-processing, such as data transmission. In addition, in a wireless sensor network environment, when a large amount of sensing data is required, each wireless sensor may cause a rapid increase in power consumption in the process of data sensing, data acquisition, data transmission, and the like. The management technology of the wireless sensor network data indicates a new direction for the research of the wireless sensor network application layer software system, and provides a new opportunity.

Disclosure of Invention

The invention aims to solve the problems of poor accuracy and real-time performance of obtained data, node energy consumption and data post-processing capability of the conventional wireless sensor network, and provides a data processing method based on a wireless sensor network in a big data environment.

In the present invention, a linear matrix inequality is used to perform matrix processing, which is introduced as follows, and in recent years, the linear matrix inequality is widely used to solve a series of problems in systems and control. With the popularization of the LMI control tool box in Matlab, the LMI tool has received attention. In practical applications, it is found that many control problems in engineering can be converted into a linear matrix inequality form, a feasibility problem or an optimization problem with various constraints is solved through the linear inequality, and the application of the LMI to solve the control problem of the system has become a great research focus in these fields. The LMI toolset provides tools for solving linear matrix inequalities, and the matrix inequalities can be described by matrix blocks through programming tools; information that can fully describe the inequality; the existing matrix inequality can be corrected; a fixed solving function is used for realizing a specific solving function; the solution results of the system can be examined.

Three linear matrix inequalities are given below. The solver is given in the LMI toolbox, wherein x represents the vector formed by the decision variables, namely the matrix variable x₁,…,x_kDirection of independent argument in (1)Amount of the compound (A).

1. Problem of feasibility

One x ∈ RⁿSuch that the LMI system: a (x) < B (x) is true,

the solver for this problem is feasp. The feasp solver is an assisted convex optimization problem by solving:

s.t.A(x)-B(x)≤tI

to solve the feasibility problem of the linear matrix inequality.

Mincx solver

The problem is the minimization of a linear objective function with linear matrix inequality constraints

s.t.A(x)＜B(x)

3. Problem of minimization of generalized eigenvalues

s.t.C(x)＜D(x)

0＜B(x)

A(x)＜λB(x)

The corresponding solver is gevp. In practical application, a suitable solver is selected according to the actual condition of the system so as to obtain the required parameters.

Definition 1: given γ > 0, if the system satisfies both of the following conditions:

(i) (mean square stability) external disturbance w_kAt 0, there is a constant φ > 0 and τ ∈ (0,1), such that

This is true.

(ii)(H_∞Performance) at zero initial conditions, for all non-zero w_k∈l₂[0, ∞) and given H_∞Performance index gamma > 0, filtering error e_kSatisfies the following conditions

The filter error system is then said to be exponential stable in the mean square sense and to have H_∞The property γ.

Lesion 1: V (η)_k) Is the Lyapunov function. If rho is more than or equal to 0, mu is more than 0, upsilon is more than 0 and 0 is more than phi and less than 1, so that

μ||η_k||²≤V_k(η_k)≤υ||η_k||²

E{V_k+1(η_k+1)|η_k}-V_k(η_k)≤ρ-φV_k(η_k)

Then there are

Theorem 2(Schurcomplement) giving a matrix of constants S₁,S₂,S₃Here, the

And

when in useWhen true, if and only if

Or

Theorem 3 let x be in the range of Rⁿ,y∈RⁿThe sum matrix Q > 0, then

x^TQy+y^TQx≤x^TQx+y^TQy

A data processing method in a big data environment based on a wireless sensor network, the method comprising:

step one, converting a plurality of control problems in engineering into a form of a linear matrix inequality, solving a feasibility problem or an optimization problem with a plurality of constraint conditions through the linear inequality, and describing a matrix processing problem as a discrete time linear system;

step two, defining variables to obtain a dimension reduction and augmentation system;

step three, designing filter parameters of the dimensionality reduction and augmentation system obtained in the step two to obtain a filter;

step four, the filter parameters are obtained by reverse deduction of the designed filter, so that the filter of the filter parameters is used for processing the big data

The signals are processed, so that the data is restored to the maximum extent, and the effectiveness of the data is ensured.

The invention has the beneficial effects that:

the discrete-time model of the system is given as:

C₁＝[18.99720.7440]C₂＝0.6，

D₁＝[-0.300.6]D₂＝0.4.

in the simulation, it is assumed that

Namely, the probability that the observation data can be received on time in the transmission process is 0.6, and the probability of one-step random time delay is 0.112. When w is_kUsing a mean of zero and a variance Q_wThe simulation results are shown in fig. 3 for 1 white noise. Figure 4 gives a standard Kalman filter estimation curve under the same conditions. It is obvious from the simulation result that the estimation algorithm provided by the invention has higher estimation precision than Kalman filtering. It follows that in a network environment, i.e. when observing numbersThe filter of the present invention is more efficient than a standard Kalman filter in the presence of random time delays.

Drawings

FIG. 1 is a block diagram of a wireless sensor network system according to the present invention;

FIG. 2 is a flow chart of an algorithm involved in the present invention;

FIG. 3 shows the true values and H according to the invention_∞Filtering the value;

fig. 4 shows the real values and Kalman filtered values to which the present invention relates.

Detailed Description

The first embodiment is as follows:

in this embodiment, a data processing method in a big data environment based on a wireless sensor network is implemented by the following steps for a wireless sensor network system structure shown in fig. 1, as shown in fig. 2:

step one, converting a plurality of control problems in engineering into a form of a linear matrix inequality, solving a feasibility problem or an optimization problem with a plurality of constraint conditions through the linear inequality, and describing a matrix processing problem as a discrete time linear system;

step two, defining variables to obtain a dimension reduction and augmentation system;

step three, designing filter parameters of the dimensionality reduction and augmentation system obtained in the step two to obtain a filter;

and step four, the filter parameters are obtained by reverse-deducing the designed filter, so that the filter of the filter parameters processes the large data signals, the data is restored to the maximum extent, and the effectiveness of the data is ensured.

The second embodiment is as follows:

different from the first embodiment, in the first embodiment, a data processing method in a big data environment based on a wireless sensor network is described, where the first step is to describe a problem as a discrete-time linear system:

wherein x is_k∈RⁿIs a vector of the states of the system,is the observed output, w_k∈R^pIs a disturbance input, z_k∈R^mIs the estimated state, A, B, C₁,C₂,D₁,D₂Is a matrix of known constants; the observed data received by the filter may have random time lag or even data loss, and the case with one-step random time lag is described as follows:

wherein, y_k∈R^rIs an observation received by the filter, ξ_i,k(i ═ 1,2) are mutually uncorrelated random sequences that satisfy the Bernoulli distribution, and satisfy the statistical probability

Wherein

When ξ_1,kWhen 1, it means that the data is received on time, and the probability is

When ξ_1,k＝0,ξ_1,k-1＝0,ξ_2,kWhen the time is 1, the time indicates that one step of random time delay exists in the data transmission process, and the probability is

From ξ_i,kCan be distributed to

The third concrete implementation mode:

different from the first or second embodiment, the data processing method in the big data environment based on the wireless sensor network of the embodiment,

in the second step, the process of defining the variables to obtain the dimension reduction and augmentation system is as follows,

definition of

Thus, can obtain

Order to

The dimension reduction and augmentation system comprises:

wherein,

and phi_i,Γ_i,i＝0,1,2,H_iI is 0,1 and

the definition is as follows:

the amplification system containing a random variable theta_i,k(i ═ 1,2), giving the following statistical properties:

the fourth concrete implementation mode:

different from the third specific embodiment, in the third step, the dimension reduction and amplification system obtained in the second step designs the filter parameters, and the process of obtaining the filter is that the filter parameter a is performed on the system_f,B_f,C_f,D_fThe filter is of the following formula:

here, the

Is an estimate of the state of the dimension-expansion,

is the state z to be estimated_kFilter of A_f,B_f,C_f,D_fIs the filter parameter to be designed; defining a filtering error asThis results in a filter error system

Wherein

F₀＝D₂,F₁＝-D_fC₂

For convenience of the following description, the following notation is introduced:

when an external disturbance w_kWhen 0, for filtering error system definition

Then there are:

where A is_1,2＝A₁-A₂,

Then another Ψ < 0 is equivalent to the inequalityIt is true that the first and second sensors,

namely, it isWherein the lambda is more than 0 and less than or equal to α_max(Ψ) there must be 0 < α ≦ ν, where λ ═ λ_max(P) then have

The system obtained according to definition 1 and lemma 1 is stable in mean square index;

when w is_kWhen the signal is not equal to 0, the signal is transmitted,

wherein

According to introduction 2

Thus, can obtain

Adding K from 0 to ∞ can obtain

At zero initial conditions V₀(η₀) When being equal to 0, then there is

The system has a given H_∞And (4) performance.

The fifth concrete implementation mode:

different from the fourth embodiment, the data processing method in the big data environment based on the wireless sensor network of the present embodiment,

in the fourth step, the process of obtaining the filter parameters by the reverse-deducing of the designed filter is as follows,

assuming the inequality of the following equation holds:

wherein,

then there isFrom the lemma 2, X-Z > 0, and the presence of the non-singular matrices R and S makes the formula true. Order to

Then there is

Wherein Q is-R^TYS^-T＝S^-1Y(X-Y^-1)YS^-TIs greater than 0. And X-RQ^-1R^T＝(I-XY)(X-Y^-1) (I-YX) > 0 from 2, P > 0.

Respectively multiplying inequalities left and right by the following matrix

diag{I,Y,I,I,Y,I,Y,I,Y,I,Y,I,I}

To obtain

Wherein

Π₁＝diag{-Π,-Π,-Π,-Π},Π₂＝diag{-I,-I}

And is

The calculation results are that:

thus, the inequality can be rewritten as follows:

on both sides of the inequality, respectively, to the left

Right passenger

Inequality equivalent to can be obtained

The above discussion has been demonstrated, and thus assuming the inequality holds, it will be assumed that the inequality is identical

Comparing to obtain the parameters of the filter to be designed

The large data signals are processed through the filter with the designed filter parameters, so that the data is restored to the maximum extent, and the effectiveness of the data is guaranteed.

Example 1:

the problem is described as a discrete-time linear system as follows:

wherein x is_k∈RⁿIs a vector of the states of the system,

is the observed output, w_k∈R^pIs an interference input and belongs to the square integrable₂[0, ∞) space, z_k∈R^mIs the estimated state, A, B, C₁,C₂,D₁,D₂Is a matrix of known constants. Observed output

The observation data received by the filter may have a step of random time delay or even data loss, and the corresponding mathematical model may be described as follows:

wherein, y_k∈R^rIs an observation received by the filter, ξ_i,k(i ═ 1,2) are mutually uncorrelated random sequences that satisfy the Bernoulli distribution, and satisfy the statistical probability:

wherein

Since the sensor and the filter communicate with each other via the network, the data transmission inevitably causes delay and packet loss, and the delay and packet loss occur randomly, here ξ_i,k(i ═ 1, 2).

When ξ_1,kWhen 1, it means that the data is received on time, and the probability is

When ξ_1,k＝0,ξ_1,k-1＝0,ξ_2,kWhen the time is 1, the time indicates that one step of random time delay exists in the data transmission process, and the probability is

When ξ_1,k＝0,ξ _1,k-11 or ξ_1,k＝0,ξ_1,k-1＝0,ξ_2,kWhen 0, it means that the packet is lost, and its probability is:

from ξ_i,kThe distribution of (c) can be found in:

E{ξ_i,k(1-ξ_i,k)}＝0,

introducing a new variable definition method, and enabling:

θ_1,k＝ξ_1,k,θ_2,k＝(1-ξ_1,k)ξ_2,k+1.

note ξ_i，k(1-ξ_i，k)＝0，ξ_i，kAndare equivalent, so there are

θ_1,kθ_2,k＝ξ_1,k(1-ξ_1,k)ξ_2,k+1＝0

From the formula:

defining:

then there are:

thus, there are obtained:

Y_k＝(θ_1,k+θ_2,k)C₁x_k+(θ_1,k+θ_2,k)C₂w_k+(1-θ₁，_k-θ_2,k)Y_k-1

y_k＝θ_1,kC₁x_k+θ_1,kC₂w_k+(1-θ_1,k)Y_k-1

order to

The following augmentation system:

wherein

And

the following can be written:

and phi_i,Γ_i,i＝0,1,2,H_iI is 0,1 and

the definition is as follows:

the system is a system containing a random variable theta_i,k(i ═ 1,2), giving the following statistical properties:

the filter is designed in the following form

Here, the

Is an estimate of the state of the dimension-expansion,

is the state z to be estimated_kFilter of A_f,B_f,C_f,D_fAre the filter parameters to be designed. Defining a filtering error as

The following two conditions are satisfied:

(i) (mean square stability) external disturbance w_kWhen 0, the constants Φ > 0 and τ ∈ (0,1) exist, so that the following expression holds.

(ii)(H_∞Performance) at zero initial conditions, for all non-zero w_k∈l₂[0, ∞) and given H_∞Performance index gamma > 0, filtering error e_kThe following are satisfied:

the filter error system is then said to be exponential stable in the mean square sense and to have H_∞The property γ.

Claims (5)

1. A data processing method based on a wireless sensor network under a big data environment is characterized in that: the method is realized by the following steps:

converting a control problem in engineering into a form of a linear matrix inequality, solving a feasibility problem or an optimization problem with multiple constraint conditions through the linear inequality, and describing a matrix processing problem as a discrete time linear system;

step two, defining variables to obtain a dimension reduction and augmentation system;

step three, designing filter parameters of the dimensionality reduction and augmentation system obtained in the step two to obtain a filter;

and step four, the filter parameters are obtained by reverse-deducing the designed filter, so that the filter of the filter parameters processes the large data signals, the data is restored to the maximum extent, and the effectiveness of the data is ensured.

2. The data processing method in the big data environment based on the wireless sensor network according to claim 1, wherein: the first step is to describe the problem as a discrete-time linear system:

wherein x is_k∈RⁿIs a vector of the states of the system,

is the observed output, w_k∈R^pIs a disturbance input, z_k∈R^mIs the estimated state, A, B, C₁,C₂,D₁,D₂Is a matrix of known constants; the observed data received by the filter may have random time lag or even data loss, and the case with one-step random time lag is described as follows:

wherein, y_k∈R^rIs an observation received by the filter, ξ_i,k(i ═ 1,2) are mutually uncorrelated random sequences that satisfy the Bernoulli distribution, and satisfy the statistical probability

Wherein

When ξ_1,kWhen 1, it means that the data is received on time, and the probability is

When ξ_1,k＝0,ξ_1,k-1＝0,ξ_2,kWhen the time is 1, the time indicates that one step of random time delay exists in the data transmission process, and the probability is

From ξ_i,kCan be distributed to

3. The data processing method in the big data environment based on the wireless sensor network according to claim 2, characterized in that: in the second step, the process of defining the variables to obtain the dimension reduction and augmentation system is as follows,

definition of theta_1,k＝ξ_1,k,

Thus, can obtain

Order to

The dimension reduction and augmentation system comprises:

wherein,

and phi_i,Γ_i,i＝0,1,2,H_iI is 0,1 and

the definition is as follows:

H₀＝[0 I 0]，H₁＝[C₁-I 0]，

the amplification system containing a random variable theta_i,k(i ═ 1,2), giving the following statistical properties:

4. the data processing method in the big data environment based on the wireless sensor network according to claim 3, wherein: in the third step, the dimension reduction and amplification system obtained in the second step designs filter parameters, and the process of obtaining the filter is that the filter parameter A is carried out on the system_f,B_f,C_f,D_fThe filter is of the following formula:

here, the

Is an estimate of the state of the dimension-expansion,

is the state z to be estimated_kFilter of A_f,B_f,C_f,D_fIs the filter parameter to be designed; defining a filtering error as

This results in a filter error system

Wherein

F₀＝D₂,F₁＝-D_fC₂

For convenience of the following description, the following notation is introduced:

when an external disturbance w_kWhen 0, for filtering error system definitionThen there are:

where A is_1,2＝A₁-A₂,

Then another Ψ < 0 is equivalent to the inequality

It is true that the first and second sensors,

namely, it is

Wherein the lambda is more than 0 and less than or equal to α_max(Ψ) there must be 0 < α ≦ ν, where λ ═ λ_max(P) then have

The obtained system is stable in mean square index;

when w is_kWhen the signal is not equal to 0, the signal is transmitted,

wherein

It can be known that

Thus, can obtain

Adding K from 0 to ∞ can obtain

At zero initial conditions V₀(η₀) When being equal to 0, then there is

The system has a given H_∞And (4) performance.

5. The data processing method in the big data environment based on the wireless sensor network according to claim 4, wherein: in the fourth step, the process of obtaining the filter parameters by the reverse-deducing of the designed filter is as follows,

assuming the inequality of the following equation holds:

wherein,

then there is

The expression is established by X-Z > 0 and the presence of non-singular matrices R and S; order to

Y＝Z^-1，

Then there is

Wherein Q is-R^TYS^-T＝S^-1Y(X-Y^-1)YS^-TIs greater than 0; and X-RQ^-1R^T＝(I-XY)(X-Y^-1) (I-YX) > 0 indicates that P > 0;

respectively multiplying inequalities left and right by the following matrix

diag{I,Y,I,I,Y,I,Y,I,Y,I,Y,I,I}

To obtain

Wherein

Π₁＝diag{-Π,-Π,-Π,-Π},Π₂＝diag{-I,-I}

And is

The calculation results are that:

G₁Σ₂＝[-D_fH₁-D_fH₁Y].

thus, the inequality can be rewritten as follows:

on both sides of the inequality, respectively, to the left

Right passenger

Inequality equivalent to can be obtained

The above discussion has been demonstrated, and thus assuming the inequality holds, it will be assumed that the inequality is identical

Comparing to obtain the parameters of the filter to be designed

The large data signals are processed through the filter with the designed filter parameters, so that the data is restored to the maximum extent, and the effectiveness of the data is guaranteed.

Images