CN107426748B - A method for multi-sensor performance estimation in wireless network control system - Google Patents

Abstract

The invention relates to the technical field of network control, and discloses a method for estimating performance of multiple sensors in a wireless network control system, which comprises the following steps: (1) acquiring sensor parameters of a wireless network control system, and establishing dynamic characteristics and a sensor measurement equation of the process; (2) connecting a plurality of sensors, and arranging two channels between every two sensors; carrying out data transmission by utilizing one of the communicated channels, and switching the channels if the channel starts to lose packets; respectively calculating estimated values and estimation error covariance matrixes of sensors at the centers of the single channel and the multiple channels; (3) and (3) comparing the transmission performance of two channels between the two sensors, calculating the estimated value and the error covariance matrix in the circulation according to the step (2), generating a comparison graph of the transmission performance of the single channel and the transmission performance of the double channel, and generating a comparison graph of a plurality of channels. The invention utilizes the Kalman filtering algorithm to ensure that the sensor is effectively scheduled in the multi-channel transmission system so as to reduce the estimation error.

Description

A method for multi-sensor performance estimation in wireless network control system

technical field

The invention relates to the technical field of network control, in particular to a multi-sensor performance estimation method in a wireless network control system.

Background technique

With the development of modern communication science and technology, wireless networked control systems (WirelessNCSs, WNCSs) are gradually replacing traditional wired networked control systems. Compared with wired networks, wireless networks transmit data through wireless media, thus eliminating the need for The complex installation and wiring process also makes the maintenance and upgrade of the system more convenient in the future, effectively saving the cost of use. Because there are nodes in WNCSs that can move in the network, and the communication range of the system is only affected by the number of nodes and the power transmitted and received, and is not controlled by wiring, this increases the mobility and expansion of the system. sex. More importantly, WNCSs can still work well in places that people can't reach or in very harsh environments.

Although the wireless networked control system has many advantages, it still faces many challenges in practical use. On the one hand, there are still unsolved problems in Networked Control Systems (NCSs) such as packet transmission delay, packet loss, clock synchronization and packet timing disorder in WNCSs. On the other hand, the fading and susceptibility to interference of wireless channels in WNCSs, the energy-saving requirements of wireless nodes (such as wireless sensors), and the limited communication channels, bandwidth, and spectrum resources have all become obstacles to the research and application of WNCSs. In addition, for the packet loss problem of wireless channels, the existing research mainly focuses on the single sensor estimation problem. This estimation mode is difficult to obtain comprehensive and stable information and the transmission distance is limited, which cannot meet the performance requirements of the continuously improving control system.

SUMMARY OF THE INVENTION

Aiming at the deficiencies in the prior art, the present invention provides a multi-sensor performance estimation method in a wireless network control system. The method of the present invention considers the multi-sensor channel switching problem in the wireless network control system, and can ensure effective performance in a multi-channel transmission system. The purpose of scheduling sensors to reduce estimation error is to complete the optimization of multi-sensor estimation performance in wireless network control systems.

In order to solve the above technical problems, the present invention is solved by the following technical solutions.

A multi-sensor performance estimation method in a wireless network control system, comprising the following steps:

(1) Obtain the sensor parameters of the wireless network control system, and establish the dynamic characteristics of the process and the sensor measurement equation:

是稳定的；Among them, A∈R ^n*n denotes the system matrix, C∈R ^m*n denotes the observation matrix with full row rank, x _k ∈ R ⁿ and y _k ∈ R ^m are real vectors, representing the state and measurement of the sensor, respectively, w _k ∈ R ⁿ and v _k ∈ R ^m are Gaussian processes with zero mean, Q is the covariance matrix of w _k , R is the covariance matrix of v _k , satisfies Q>0 and R>0, m, n are positive integers , k represents the moment, (A, C) is detectable,

is stable;

(2) Connect multiple sensors, set two channels between each two sensors, and set the parameters of each channel at the same time; use one of the connected channels for data transmission, if the channel starts to lose packets, then the channel Switch; set the number of cycles, initialize x ₀ , P ₀ , and use the Kalman filter algorithm to calculate the estimated value and estimated error covariance matrix of the sensor at the center of the single-channel and multi-channel respectively. The calculation equation is:

为中心处传感器的估计值，P_κ|κ为估计误差协方差矩阵，计算得出P_k|k的极限值，直到循环完成；in

is the estimated value of the sensor at the center, P _κ|κ is the estimated error covariance matrix, and the limit value of P _k|k is calculated until the cycle is completed;

(3) Compare the transmission performance between the two sensors through the two channels, and generate a comparison drawing of multiple channels according to the estimated value and the estimated error covariance matrix data in the channel completion cycle in step (2).

Preferably, in step (1), the initial state x ₀ of the system is a Gaussian random vector with a mean value of 0 and a covariance matrix of P ₀ >0, and w _k , v _k and x ₀ are independent of each other.

The present invention has significant technical effects due to the adoption of the above technical solutions:

The method of the invention first establishes the dynamic characteristics of the process and the sensor measurement equation, and obtains the system parameters; then calculates the critical packet acceptance probability, that is, the error will diverge with the passage of time when the packet acceptance rate is smaller than the critical value; and then iteratively Solve the estimated value and estimated error covariance matrix of the sensor at the center of single-channel and multi-channel; finally, according to the calculated estimated value and estimated error, it is determined that multi-channel handover transmission is better than single-channel transmission, and the handover method is given to complete wireless network control Optimization of multi-sensor estimation performance in the system. The present invention utilizes the Kalman filter algorithm to design a multi-sensor performance estimation method in a wireless network control system, and can ensure effective scheduling of sensors in a multi-channel transmission system to reduce estimation errors.

Description of drawings

1 is a schematic diagram of a model structure in a multi-sensor performance estimation method in a wireless network control system of the present invention;

2 is a schematic diagram of a workflow in a method for estimating performance of multiple sensors in a wireless network control system according to the present invention;

3 is a comparison diagram of the real value and the estimated value of a single channel in a multi-sensor performance estimation method in a wireless network control system of the present invention;

FIG. 4 is a comparison diagram of estimation errors of single-channel sensor 1 and sensor 2 in a multi-sensor estimation performance method in a wireless network control system of the present invention;

5 is a graph of the average estimation error of the single-channel sensor 2 in the estimation performance method of the multi-sensor estimation performance method in a wireless network control system of the present invention;

6 is a comparison diagram of the real value and the estimated value of dual channels in a multi-sensor performance estimation method in a wireless network control system of the present invention;

7 is a comparison diagram of the estimation errors of dual-channel sensor 1 and sensor 2 in a multi-sensor estimation performance method in a wireless network control system of the present invention;

FIG. 8 is a graph of the average estimation error of dual-channel sensors 2 in a multi-sensor performance estimation method in a wireless network control system of the present invention.

Detailed ways

The present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

Example 1

As shown in Figures 1 to 8, a method for estimating performance of multiple sensors in a wireless network control system includes the following steps:

(1) Obtain the sensor parameters of the wireless network control system, and establish the dynamic characteristics of the process and the sensor measurement equation:

是稳定的；Among them, A∈R ^n*n denotes the system matrix, C∈R ^m*n denotes the observation matrix with full row rank, x _k ∈ R ⁿ and y _k ∈ R ^m are real vectors, representing the state and measurement of the sensor, respectively, w _k ∈ R ⁿ and v _k ∈ R ^m are Gaussian processes with zero mean, Q is the covariance matrix of w _k , R is the covariance matrix of υ _k , satisfies Q>O and R>0; the initial state of the system x ₀ is a Gaussian random vector with mean 0 and covariance matrix P ₀ > 0, w _k , v _k and x ₀ are independent of each other, m, n are positive integers, k represents time, (A, C) is detectable of,

is stable;

(2) Connect multiple sensors, set two channels between each two sensors, and set the parameters of each channel at the same time; use one of the connected channels for data transmission, if the channel starts to lose packets, then the channel Switch; set the number of cycles, initialize x ₀ , P ₀ , and use the Kalman filter algorithm to calculate the estimated value and estimated error covariance matrix of the sensor at the center of the single-channel and multi-channel respectively. The calculation equation is:

为中心处传感器的估计值，P_k|k为估计误差协方差矩阵，计算得出P_k|k的极限值，直到循环完成；in

is the estimated value of the sensor at the center, P _k|k is the estimated error covariance matrix, and the limit value of P _k|k is calculated until the cycle is completed;

(3) Compare the transmission performance between the two sensors through the two channels, and generate a comparison drawing of multiple channels according to the estimated value and the estimated error covariance matrix data in the channel completion cycle in step (2).

For multi-sensors in a single-channel wireless network control system, data packets may be lost during channel propagation due to channel attenuation or congestion. Similarly, γ _k indicates whether the estimator has received the measurement y _k , and the estimator also knows the information of γ _k , regardless of other uncertain factors such as delay, according to the optimality of the Kalman filter, we get:

P_k+1|k＝AP_k|kA′+Q和

P_0|-1＝P₀。Among them, K _k =P _k|k-1 C′(CP _k|k-1 C′+R) ⁻¹ , and the time update of Kalman filter is also optimal, that is,

P _k+1|k =AP _k|k A′+Q sum

P _0|-1 =P ₀ .

Due to the random data packet loss in the network, the covariance matrix of the estimated error is also random, which is different from the standard Kalman filter, and its estimated error covariance matrix is deterministic. Let P _k =P _k|k-1 , then the update equation of P _k is:

P _k+1 =AP _k A'+Q-γ _k AP _k C'(CP _k C'+R) ^-1 CP _k A'

Now output a value from the system. After reaching sensor 1, it will reach sensor 2. There are packet loss rates and delays. Suppose that from the 9th time, the sensor 1 receives the signal and successfully transmits it to 2, the 10th time. If the time transmission is unsuccessful, the 11 time is unsuccessful, and the 12 time is successful, the following relationship can be obtained, namely,

如果不成功，则由上一时刻传感器2已获得的值进行计算。可以得到，在单信道无线网络控制系统中多传感器的估计误差协方差很大概率上将以指数型增长。As can be seen from the figure, when sensors 1 and 2 are successful, the value obtained by sensor 2 is 1 and the output value is multiplied by the system coefficient A, and the success probability is

If unsuccessful, it is calculated from the value obtained by sensor 2 at the last moment. It can be obtained that the estimation error covariance of multiple sensors in a single-channel wireless network control system will increase exponentially with a high probability.

FIG. 1 shows a model diagram of a dual-channel wireless network control system.

Figure 2 presents the flow chart of the method for multi-sensor performance estimation in the wireless network control system. Specifically, it can be described as follows:

A multi-sensor performance estimation method in a wireless network control system, the method comprising the following steps:

Step 1: Establish process dynamics and sensor measurement equations:

可控，(A，C)可观，量测矩阵C是行满秩的，即rank(C)＝m≤n，系统的初始状态x₀是均值为0和协方差矩阵为P₀＞0的高斯随机向量，进一步，w_k，v_k和x₀是相互独立的。where A∈R ^n*n denotes the system matrix, C∈R ^m*n denotes the observation matrix, x _k ∈ R ⁿ and y _k ∈ R ^m denote the state and measurement of the system, respectively, w _k ∈ R ⁿ and v _k ∈R _m is a Gaussian process with zero mean, and its covariance matrices are Q>0 and R>0, respectively.

Controllable, (A, C) is observable, the measurement matrix C is full rank, that is, rank(C)=m≤n, the initial state x ₀ of the system is the mean value of 0 and the covariance matrix of P ₀ >0 Gaussian random vector, further, w _k , v _k and x ₀ are independent of each other.

Step 2: The measurement y _k of the system is transmitted to the remote estimator through an unreliable communication channel. Due to channel attenuation or congestion, data packets may be lost during channel propagation. Intuitively, the greater the probability of packet loss and the more serious the loss of information, the greater the possibility of divergence of the estimated error covariance matrix P _k|k-1 . For the independent and identically distributed packet loss process, there is a critical probability of receiving packets critical so that as long as the probability of receiving packets is greater than this value, the estimated error covariance matrix can achieve average boundedness.

First find the eigenvalues of matrix A:

eigvalues=eig(A)

Then calculate the spectral radius of matrix A:

spectrumofA=max(abs(eigvalues))

Finally, the critical acceptance probability is obtained:

Step 3: Using the Kalman filter algorithm, calculate the estimated value and estimated error covariance matrix of the sensor at the center of the single-channel and multi-channel respectively:

Standard Kalman filter equation:

It can be obtained from the above formula:

This gives the limit of P _k|k , which satisfies the equation X=g(h(X)). For multi-sensors in a single-channel wireless network control system, data packets may be lost during channel propagation due to channel attenuation or congestion. It can be obtained that the estimation error covariance of multi-sensors in a single-channel wireless network control system is very high. There is a high probability that it will grow exponentially.

Step 4: Now compare the transmission performance comparison between the two sensors over the two channels:

Assuming channel 1:

Assuming channel 2:

The reason why the transition probabilities of the two channels are defined, namely T ₁ (1, 1)>0.5, T ₁ (2, 2)>0.5, T ₂ (1, 1)>0.5, T ₂ (2, 2)>0.5 is that Because such channels are called Gilbert-Elliott channels, the memory that such channels generally exist depends on the transition probabilities between states. Markov proposed that in the process of transferring certain factors of a system, the Nth result only depends on the influence of the N-1th result, that is, it is only related to the current state and has nothing to do with the previous state, so when T ₁ (1, 1 )>0.5, when T ₁ (2, 2)>0.5, the packet acceptance rate of the two elements of the transition matrix at the next moment is still high to avoid errors.

Taking two channels as an example, it is assumed that the sensors first transmit through channel 1, and then fail to transfer to channel 2 after n1 successes, and then fail to switch back to channel 1 after n2 successes. The reason for this scheduling method is that in the first After a channel transmission error, the system begins to generate errors. If it continues to transmit through this channel, its packet loss rate is greater than the packet reception rate, which is easy to cause the error value to start to increase exponentially. Exponential growth of this error is avoided. Similarly, it can be extended to multi-channel systems.

Step 5: According to the estimated value obtained by calculation, it is determined that multi-channel switching transmission is better than single-channel transmission, and a channel switching method is given to complete the optimization of multi-sensor estimation performance in the wireless network control system.

Further, the channel switching method obtained in the step 4 can ensure effective scheduling of sensors in a multi-channel transmission system to reduce estimation errors.

The technical solutions of the present invention will be further elaborated below through specific examples. In the experiment, single-channel and dual-channel wireless network control systems are used to compare and verify the algorithm. Specifically, the following experimental parameters were used:

观测矩阵C＝[3 2]，系统协方差矩阵

观测协方差矩阵R＝0.6，设置循环次数为1000，取0～1之间的随机数作为估计值x_k|k、估计误差协方差矩阵P_k|k和真实值x_k的初始，定义初始观测值y_k＝0，真实值x_k第二列为[1 0.4]，设置信道参数

Assume the system matrix separately

Observation matrix C=[3 2], system covariance matrix

The observation covariance matrix R=0.6, the number of cycles is set to 1000, the random number between 0 and 1 is taken as the initial value of the estimated value x _k|k , the estimated error covariance matrix P _k|k and the real value x _k , and the initial value is defined. The observed value y _k =0, the second column of the true value x _k is [1 0.4], and the channel parameters are set

如果不成功，则由上一时刻传感器2已获得的值进行计算。可以得到，在单信道无线网络控制系统中多传感器的估计误差协方差将以指数型增长；对于双信道系统，假设传感器之间先通过信道1传输，成功n1次后失败转换为信道2传输，再成功n2次后失败换回信道1，之所以采用这样的调度方式是因为在第一个信道传输错误后系统开始产生误差，倘若继续由该信道传输则它的丢包率更大于收包率，容易使误差值开始以指数型增长，这时候换了一个信道系统重新开始传输则避免了这种误差的指数型增长。Kalman filtering is performed on single-channel and dual-channel wireless network control systems. For a single-channel system, a value is output, and after reaching sensor 1, it reaches sensor 2. There are packet loss rates and delays, and the success between sensors 1 and 2 is successful. When the value obtained by sensor 2 is 1, the output value is multiplied by the system coefficient A, and the success probability is

If unsuccessful, it is calculated from the value obtained by sensor 2 at the last moment. It can be obtained that the estimated error covariance of multi-sensors in a single-channel wireless network control system will increase exponentially; for a dual-channel system, it is assumed that the sensors are transmitted through channel 1 first, and after successful n1 times, the failure is converted to channel 2 transmission. After successful n2 times, it fails to switch back to channel 1. The reason for this scheduling method is that the system starts to generate errors after the first channel transmission error. If the channel continues to transmit, its packet loss rate is greater than the packet reception rate. , it is easy to make the error value start to increase exponentially. At this time, changing a channel system and restarting transmission avoids the exponential increase of this error.

Figures 3, 4, 5, 6, 7, and 8 are the results of simulation verification of the designed method through the Matlab system.

Figure 3 shows the comparison of the true value, the estimated value at sensor 1 and the estimated value at sensor 2 in a single-channel system, and Figure 4 shows the estimated error at sensor 1 and sensor 2 in a single-channel system. The estimation error comparison graph of , Figure 5 presents the average estimation error at sensor 2 in a single-channel system. It can be seen from the figure that the estimation error and the average estimation error at sensor 2 are relatively large.

Figure 6 shows the comparison of the true value, the estimated value at sensor 1 and the estimated value at sensor 2 in a dual-channel system, and Figure 7 shows the estimated error at sensor 1 and sensor 2 in a dual-channel system. The estimation error comparison graph of , Figure 8 presents the average estimation error at sensor 2 in a two-channel system. It can be seen from the figure that the estimation error and the average estimation error at sensor 2 are relatively small.

The method of the invention first establishes the dynamic characteristics of the process and the sensor measurement equation to obtain the system parameters; then calculates the critical packet acceptance probability, that is, the error will diverge with the passage of time when the packet acceptance rate is smaller than the critical value; and then iteratively Solve the estimated value and estimated error covariance matrix of the sensor at the center of single-channel and multi-channel; finally, according to the calculated estimated value and estimated error, it is determined that multi-channel switching transmission is better than single-channel transmission, and the switching method is given to complete the wireless network control Optimization of multi-sensor estimation performance in the system. The present invention utilizes the Kalman filter algorithm to design a multi-sensor performance estimation method in a wireless network control system, and can ensure effective scheduling of sensors in a multi-channel transmission system to reduce estimation errors.

In a word, the above are only preferred embodiments of the present invention, and all equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope of the patent of the present invention.

Claims (2)

1. a multi-sensor performance estimation method in a wireless network control system, is characterized in that, comprises the steps: (1) Obtain the sensor parameters of the wireless network control system, and establish the dynamic characteristics of the process and the sensor measurement equation:

Among them, A∈R ^n*n denotes the system matrix, C∈R ^m*n denotes the observation matrix with full row rank, x _k ∈ R ⁿ and y _k ∈ R ^m are real vectors, representing the state and measurement of the sensor, respectively, w _k ∈ R ⁿ and v _k ∈ R ^m are Gaussian processes with zero mean, Q is the covariance matrix of w _k , R is the covariance matrix of υ _k , satisfies Q>0 and R>0, m, n are positive integers , k represents the moment, (A, C) is detectable,

is stable; (2) Connect multiple sensors, set two channels between each two sensors, and set the parameters of each channel at the same time; use one of the connected channels for data transmission, if the channel starts to lose packets, then the channel Switch; set the number of cycles, initialize x ₀ , P ₀ , and use the Kalman filter algorithm to calculate the estimated value and estimated error covariance matrix of the sensor at the center of the single-channel and multi-channel respectively. The calculation equation is:

in

is the estimated value of the sensor at the center, P _k|k is the estimated error covariance matrix, and the limit value of P _k|k is calculated until the cycle is completed; (3) Fix two adjacent sensors, calculate the estimated value and error covariance matrix in the cycle according to step (2), and generate a comparison chart of single-channel and dual-channel transmission performance. 2. The multi-sensor performance estimation method in a wireless network control system according to claim 1, characterized in that: in step (1), the initial state x ₀ of the system is that the mean value is 0 and the covariance matrix is P ₀ > A Gaussian random vector of 0, w _k , v _k and x ₀ are independent of each other.

Images