阅读说明：本技术 一种室内移动目标定位方法、装置和计算机设备 (Indoor mobile target positioning method and device and computer equipment ) 是由 曹宁 马利 毛明禾 于 2021-08-31 设计创作，主要内容包括：本发明公开了一种室内移动目标定位方法,包括：建立未知移动目标节点运动的状态方程和观测模型；根据状态方程,通过UT计算的一组Sigma点集及其对应的权值,获得系统状态量的一步预测值及协方差矩阵、和观测Sigma点集；根据观测模型,计算观测Sigma点集的观测预测值,得到系统的观测预测值；根据基于最大相关熵的代价函数,融合系统的观测预测值,获得重构观测模型的观测协方差和观测互协方差矩阵,确定卡尔曼增益矩阵,估计系统状态量,确定未知移动目标节点的位置坐标。本发明中基于MCUKF的室内定位方法能更好地完成量测噪声出现非高斯噪声情况下的室内定位,定位精度相比于LS、EKF和UKF的室内定位方法较高,以及鲁棒性比LS、EKF和UKF的室内定位方法也较强。(The invention discloses a method for positioning an indoor moving target, which comprises the following steps: establishing a state equation and an observation model of the motion of an unknown moving target node; according to a state equation, obtaining a one-step predicted value and a covariance matrix of the system state quantity and an observation Sigma point set through a group of Sigma point sets calculated by UT and corresponding weights thereof; calculating an observation predicted value of observing a Sigma point set according to the observation model to obtain an observation predicted value of the system; and according to a cost function based on the maximum correlation entropy, fusing an observation predicted value of the system, obtaining an observation covariance matrix and an observation cross covariance matrix of a reconstructed observation model, determining a Kalman gain matrix, estimating a system state quantity, and determining the position coordinate of an unknown moving target node. The indoor positioning method based on the MCUKF can better complete indoor positioning under the condition that non-Gaussian noise occurs in measurement noise, the positioning precision is higher than that of the indoor positioning methods of LS, EKF and UKF, and the robustness is also higher than that of the indoor positioning methods of LS, EKF and UKF.)

1. An indoor mobile target positioning method comprises the following steps: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3; the positioning method is characterized by comprising the following steps:

establishing a state equation of the motion of an unknown moving target node;

establishing an observation model of unknown moving target node motion;

according to a state equation, obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by unscented transformation UT and corresponding weights thereof; according to the predicted value of the first step, calculating a new Sigma point set again through unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set;

calculating an observation predicted value of an observation Sigma point set according to the observation model, and obtaining the observation predicted value of the system through weighted summation;

according to a cost function based on the maximum correlation entropy, fusing an observation predicted value of the system to obtain an observation covariance matrix and an observation cross covariance matrix of a reconstructed observation model; determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix;

estimating a system state quantity and a covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the estimated system state quantity and the covariance matrix.

2. The indoor moving-object positioning method of claim 1, wherein the establishing of the state equation of the unknown moving-object node motion comprises:

taking the position and the speed of an unknown mobile target node as state quantitiesEstablishing a state equation of the motion of an unknown moving target node; and the motion of the unknown moving target node is described by the change of x (k) at time step k, and the expression is as follows: x (k +1) ═ f (x (k)) + ω (k)

Wherein (x, y) is the spatial location coordinates of the unknown node;the speed of the unknown node in the x direction and the y direction respectively; f (x (k)) is the system matrix,w (k) is process noise, and the sampling time interval is set as T ═ 1 s; the abbreviation is: x (k +1) ═ f (x (k)) + w (k).

3. An indoor moving-object positioning method as claimed in claim 2, wherein said establishing an observation model of unknown moving-object node motion comprises:

acquiring an RSSI value of an unknown mobile target node received from a known wireless Access Point (AP) node;

and establishing an observation model by taking the RSSI value as an observed quantity.

4. The indoor moving-object positioning method of claim 3, wherein the establishing an observation model using the RSSI value as an observed quantity comprises:

taking the weighted average of RSSI values of different known wireless access point AP nodes as an observed quantity, and establishing an observation model by adopting a logarithmic distance path loss model; and the expression of the observation model is as follows:

Z(k)＝h(x(k))+v(k)

wherein the logarithmic distance path loss model has a general formula:rs (d) is the path loss in dB for a distance d between the transceivers; RS (d)₀) Is a short distance d₀Reference path loss of time, d₀Taking 1 meter; λ is a loss factor;is a normal random variable;v (k) is the observation noise,is the coordinate vector, X, of the Lth known wireless access point AP node_kIs the coordinate vector of the unknown moving target node at the kth moment.

5. The indoor moving object positioning method of claim 4, wherein the obtaining an observed Sigma point set comprises:

a set of sampling points, called the Sigma point set, is obtained from the UT transform:

calculating the corresponding weight of the sampling point as follows:

form 2n +1 Sigma point sets according to the system equation:

the predicted value and covariance matrix of the system state quantity are:

from the one-step predicted values, the UT transform is again used to generate a new Sigma point set:

substituting the predicted new Sigma point set into the observation model to obtain a corresponding observed Sigma point set:

wherein n is the dimension of the state; x⁰The sampling point when i is 0; xⁱIs the ith Sigma sample point;is the sample mean of the Sigma set; p is the variance of the state vector; λ is a scaling parameter;weight values of the covariance and the mean, respectively; alpha is a distribution state parameter of the sampling point; beta is more than or equal to 0 and is a non-negative weight coefficient; xⁱ(k +1 | k) is a one-step predicted value of the Sigma sampling point;and P (k +1 |. k) is a one-step predicted value and covariance matrix of the system state quantity, respectively, and Q (k) is process noise; zⁱ(k +1 | k) is the observed Sigma sample point.

6. The indoor moving object positioning method of claim 5, wherein the system's observation prediction is as follows:

7. the indoor moving object locating method of claim 6, wherein the determining a Kalman gain matrix comprises:

reconstructing the observation model to obtain the following nonlinear model:

wherein the content of the first and second substances,

E[r(k)r^T(k)]＝R(k)

left-multiplying both sides of the reconstructed observation model by S^-1(k) Due to S^-1(k)＝S^T(K) The following equation is obtained:

wherein the content of the first and second substances,e(k)＝S^T(k) phi (k) then, e (k) the element e of the jth line_j(k)，e_j(k)＝d_j(k)-m_j(x(k))。

According to the nature of the correlation entropy, the cost function based on the correlation entropy is:

when the cost function J (-) is maximum, the optimal estimate of the state is obtained, then the optimal estimate of x (k) is:

order toThe above formula yields:

this gives:

the modified covariance was found to be:

wherein the content of the first and second substances,

Φ(k)＝diag(G_γ(e₁(k),…,G_γ(e_n+m(k))))

the actual state is not known in practice, letNamely, it isThen there is

The modified observed covariance was:

then the covariance of the system's observationsThe following steps are changed:

cross covariance matrixComprises the following steps:

the Kalman gain matrix K (k) is:

8. the method of claim 7, wherein updating the system state quantities and covariance matrix based on the Kalman gain matrix comprises:

computing state updatesAnd a covariance update P (k +1 |. k + 1);

9. an indoor mobile target positioning device comprising: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3; characterized in that, the positioner still includes:

the state equation establishing module is used for establishing a state equation of the motion of the unknown moving target node;

the observation model establishing module is used for establishing an observation model of unknown moving target node motion;

the observation Sigma point set determining module is used for obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by the unscented transformation UT and corresponding weights thereof according to the state equation; according to the predicted value of the first step, calculating a new Sigma point set again through unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set;

the system observation predicted value determining module is used for calculating the observation predicted value of the observation Sigma point set according to the observation model and obtaining the observation predicted value of the system through weighted summation;

the Kalman gain matrix determination module is used for fusing the observation predicted value of the system according to the cost function based on the maximum correlation entropy determined by the property of the correlation entropy to obtain an observation covariance matrix and an observation cross covariance matrix for reconstructing an observation model; determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix;

the moving target determining module is used for updating the system state quantity and the covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the updated system state quantity and the covariance matrix.

10. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor when executing the computer program implements the steps of the method of any of claims 1-8.

Technical Field

The invention relates to the field of indoor positioning, in particular to an indoor moving target positioning method, device and computer equipment based on maximum correlation entropy unscented Kalman Filter (MCUKF).

Background

In recent years, indoor positioning is increasingly demanded. Due to the application scene of huge indoor positioning scale and high commercial value, the indoor positioning system is widely applied to the fields of medical centers, intelligent homes, shopping centers, positioning and tracking of personnel in mines, tracking of goods and the like. Indoor positioning systems based on RSS have received much attention as soon as they are available, due to the advantage that they do not require specific hardware. As Kalman filtering based on MMSE (minimum mean square error) criterion is an optimal state estimation method, the method is also introduced into indoor positioning and achieves better effect. Besides, some modified algorithms of kalman filtering, such as Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), etc., are also applied to the indoor positioning system. However, in an actual indoor environment, due to a very complex environment, a wireless signal is prone to phenomena such as reflection, diffraction and refraction, and meanwhile, changes of surrounding environments such as temperature, humidity and obstacles and non-line-of-sight and the like can also have certain influences on propagation of the wireless signal, so that the indoor positioning system is often influenced by multimodal heavy-tailed non-gaussian noise due to the factors, the accuracy of the kalman filter based on the MMSE criterion is greatly reduced, and even filtering divergence is caused, so that the indoor positioning system cannot be normally used.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a method, an apparatus and a computer device for positioning an indoor moving target.

The embodiment of the invention provides a method for positioning an indoor moving target, which comprises the following steps: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3; the positioning method comprises the following steps:

establishing a state equation of the motion of an unknown moving target node;

establishing an observation model of unknown moving target node motion;

according to a state equation, obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by unscented transformation UT and corresponding weights thereof; according to the predicted value of the first step, calculating a new Sigma point set again through unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set;

calculating an observation predicted value of an observation Sigma point set according to the observation model, and obtaining the observation predicted value of the system through weighted summation;

according to a cost function based on the maximum correlation entropy, fusing an observation predicted value of the system to obtain an observation covariance matrix and an observation cross covariance matrix of a reconstructed observation model; determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix;

estimating a system state quantity and a covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the estimated system state quantity and the covariance matrix.

Further, the establishing of the state equation of the unknown moving target node motion includes:

taking the position and the speed of an unknown mobile target node as state quantitiesEstablishing a state equation of the motion of an unknown moving target node; and the motion of the unknown moving target node is described by the change of x (k) at time step k, and the expression is as follows: x (k +1) ═ f (x (k)) + ω (k)

Wherein (x, y) is the spatial location coordinates of the unknown node;the speed of the unknown node in the x direction and the y direction respectively; f (x (k)) is the system matrix,w (k) is process noise, and the sampling time interval is set as T ═ 1 s; the abbreviation is: x (k +1) ═ f (x (k)) + w (k).

Further, the establishing an observation model of the unknown moving target node motion includes:

acquiring an RSSI value of an unknown mobile target node received from a known wireless Access Point (AP) node;

and establishing an observation model by taking the RSSI value as an observed quantity.

Further, the establishing an observation model by using the RSSI value as an observed quantity includes:

taking the weighted average of RSSI values of different known wireless access point AP nodes as an observed quantity, and establishing an observation model by adopting a logarithmic distance path loss model; and the expression of the observation model is as follows:

Z(k)＝h(x(k))+v(k)

wherein the logarithmic distance path loss model has a general formula: RS (d) ═ RS (d)₀)+10λlg(d/d₀)+θ_dB(ii) a Rs (d) is the path loss in dB for a distance d between the transceivers; RS (d)₀) Is a short distance d₀Reference path loss of time, d₀Taking 1 meter; λ is a loss factor; theta_dBIs a normal random variable;v (k) is the observation noise,is the coordinate vector, X, of the Lth known wireless access point AP node_kIs the coordinate vector of the unknown moving target node at the kth moment.

Further, the obtaining a set of observed Sigma points comprises:

a set of sampling points, called the Sigma point set, is obtained from the UT transform:

calculating the corresponding weight of the sampling point as follows:

form 2n +1 Sigma point sets according to the system equation:

the predicted value and covariance matrix of the system state quantity are:

from the one-step predicted values, the UT transform is again used to generate a new Sigma point set:

substituting the predicted new Sigma point set into the observation model to obtain a corresponding observed Sigma point set:

wherein n is the dimension of the state; x⁰The sampling point when i is 0; xⁱIs the ith Sigma sample point;is the sample mean of the Sigma set; p is the variance of the state vector; λ is a scaling parameter;weight values of the covariance and the mean, respectively; alpha is a distribution state parameter of the sampling point; beta is more than or equal to 0 and is a non-negative weight coefficient; xⁱ(k +1 | k) is a one-step predicted value of the Sigma sampling point;and P (k +1 |. k) is a one-step predicted value and covariance matrix of the system state quantity, respectively, and Q (k) is process noise; zⁱ(k +1 | k) is the observed Sigma sample point.

Further, the system's observed predictions are as follows:

further, the determining a kalman gain matrix includes:

reconstructing the observation model to obtain the following nonlinear model:

wherein the content of the first and second substances,

E[r(k)r^T(k)]＝R(k)

left-multiplying both sides of the reconstructed observation model by S^-1(k) Due to S^-1(k)＝S^T(K) The following equation is obtained:

wherein the content of the first and second substances,e(k)＝S^T(k)φ(k)

then element e of line j in e (k)_j(k)，e_j(k)＝d_j(k)-m_j(x(k))。

According to the nature of the correlation entropy, the cost function based on the correlation entropy is:

when the cost function J (-) is maximum, the optimal estimate of the state is obtained, then the optimal estimate of x (k) is:

order toThe above formula yields:

this gives:

the modified covariance was found to be:

wherein the content of the first and second substances,

Φ(k)＝diag(G_Υ(e₁(k),…,G_Υ(e_n+m(k))))

the actual state is not known in practice, letNamely, it isThen there is

The modified observed covariance was:

then the covariance of the system's observationsThe following steps are changed:

cross covariance matrixComprises the following steps:

the Kalman gain matrix K (k) is:

further, the updating the system state quantity and the covariance matrix according to the kalman gain matrix includes:

computing state updatesAnd a covariance update P (k +1 |. k + 1);

an embodiment of the present invention further provides an indoor mobile target positioning device, including: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3; the positioning device further comprises:

the state equation establishing module is used for establishing a state equation of the motion of the unknown moving target node;

the observation model establishing module is used for establishing an observation model of unknown moving target node motion;

the observation Sigma point set determining module is used for obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by the unscented transformation UT and corresponding weights thereof according to the state equation; according to the predicted value of the first step, calculating a new Sigma point set again through unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set;

the system observation predicted value determining module is used for calculating the observation predicted value of the observation Sigma point set according to the observation model and obtaining the observation predicted value of the system through weighted summation;

the Kalman gain matrix determination module is used for fusing the observation predicted value of the system according to the cost function based on the maximum correlation entropy determined by the property of the correlation entropy to obtain an observation covariance matrix and an observation cross covariance matrix for reconstructing an observation model; determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix;

the moving target determining module is used for updating the system state quantity and the covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the updated system state quantity and the covariance matrix.

An embodiment of the present invention further provides a computer device, including: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3;

establishing a state equation of the motion of an unknown moving target node;

establishing an observation model of unknown moving target node motion;

according to a state equation, obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by unscented transformation UT and corresponding weights thereof; according to the predicted value of the first step, calculating a new Sigma point set again through unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set;

calculating an observation predicted value of an observation Sigma point set according to the observation model, and obtaining the observation predicted value of the system through weighted summation;

according to a cost function based on the maximum correlation entropy, fusing an observation predicted value of the system to obtain an observation covariance matrix and an observation cross covariance matrix of a reconstructed observation model; determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix;

updating a system state quantity and a covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the updated system state quantity and the covariance matrix.

Compared with the prior art, the indoor positioning method, the indoor positioning device and the computer equipment provided by the embodiment of the invention have the following beneficial effects:

aiming at the problem that an indoor positioning system is generally influenced by multimodal heavy-tailed non-Gaussian noise in a complex indoor environment, the invention provides an indoor positioning method based on maximum cross-correlation entropy criterion (MCC) and an Unscented Kalman Filter (UKF), namely based on maximum cross-correlation entropy unscented Kalman filter (MCUKF). In addition, indoor positioning experiments are carried out on an indoor positioning method based on LS, EKF and UKF and an indoor positioning method based on MCUKF, and the results show that: the indoor positioning precision of the indoor positioning method based on LS, EKF and UKF is lower than that of the MCUKF method of the design; meanwhile, an indoor positioning method based on MCUKF under four different nuclear widths is analyzed, and the effect of the proposed method is optimal when upsilon ═ 10 is obtained. The indoor positioning method based on the MCUKF can better complete indoor positioning under the condition that non-Gaussian noise occurs in measurement noise, the positioning precision is higher than that of the indoor positioning methods of LS, EKF and UKF, and the robustness is also higher than that of the indoor positioning methods of LS, EKF and UKF. Namely, the MCUKF has better estimation effects on the state estimation and the position positioning of the unknown node than the traditional Extended Kalman Filtering (EKF) and Unscented Kalman Filtering (UKF) and has stronger robustness in the environment of multimodal heavy-tailed non-Gaussian observation noise.

Drawings

FIG. 1 is a flow diagram of a method for locating an indoor moving object provided in one embodiment;

FIG. 2 is a comparison graph of positioning accuracy for different positioning algorithms provided in one embodiment;

FIG. 3 is a comparison of MCUKF algorithm dynamic positioning for different kernel widths as provided in one embodiment;

FIG. 4 is a comparison graph of localized RMSE at different core widths as provided in one example.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

In one embodiment, referring to fig. 1, a method for positioning an indoor moving target is provided, which specifically includes:

the MCUKF indoor positioning method reconstructs observation information based on a non-linear regression method of MCC, and enhances the robustness of the UKF to non-Gaussian noise.

Setting L wireless Access Point (AP) nodes in an indoor field, wherein the coordinate of the ith AP node is (x)_l,y_l) And L is 1,2, … L. The moving unknown node can receive the signal strength from the AP node and takes the position and the speed of the moving unknown node as state quantities(x, y) are the spatial unknown coordinates of the unknown node,the velocity of the unknown node in the x direction and the y direction respectively is described by the change of x (k) at the time step k, and the expression is as follows: x (k +1) ═ f (x (k)) + ω (k)

Wherein the content of the first and second substances,w (k) is process noise, and the sampling time interval is assumed to be T ═ 1 s.

It can also be abbreviated as: x (k +1) ═ f (x (k)) + w (k).

The RSSI ranging technique is also called signal strength ranging, and is calculated based on a radio propagation path loss model. In such a complicated and changeable indoor environment, traditional wireless signal transmission models such as a free space model are no longer applicable, and the current more common models are a logarithmic distance path loss model and the like, which are verified in various practical environments, and the general models are as follows:

RS(d)＝RS(d₀)+10λlg(d/d₀)+θ_dB

in indoor positioning, RSSI values may be extracted from the transceivers of the mobile nodes, and the observed model of predicted RSSI values employs a logarithmic range path loss model (i.e., the RSSI values are predicted using the logarithmic range path loss model to form the observed model of RSSI values). In order to enhance the anti-interference capability of the positioning system to other external wireless signals, the obtained weighted average of the signal strength of multiple measurements of different known nodes is obtainedAs an observed quantity.

Z(k)＝h(x(k))+v(k)

Wherein the content of the first and second substances,v (k) is the observation noise,is the coordinate vector of the Lth known node, X_kIs the coordinate vector of the unknown node moving at the kth time.

It should be noted that, in the conventional observation model, it is assumed that the noise is gaussian distributed, and now the noise changes, and is a non-gaussian noise with multiple peaks and heavy tails, and needs to be reconstructed; 2. the traditional observation model is constructed based on the criterion of minimum mean square error; the current criterion is changed by adopting the maximum cross-correlation entropy criterion, the criterion is changed, and the observation model needs to be reconstructed if the originally established observation model is not consistent.

The maximum correlation entropy unscented Kalman filtering MCUKF positioning algorithm comprises the following specific processes:

(1) first, a set of sampling points (called Sigma point set) and their corresponding weights are obtained according to UT transform:

(initial value)

(2) The corresponding weights of these samples transformed by UT are calculated as follows:

(3) then 2n +1 Sigma point sets are formed according to the system equation:

(4) the one-step prediction and covariance matrix of the system state quantity are as follows:

(5) the UT transform is then used again to generate a new Sigma point set based on the predicted values from the previous step.

(6) Substituting the predicted new Sigma point set into the observation equation to obtain the corresponding observed Sigma point set

(7) Obtaining an observation predicted value of an observation Sigma point set in the step (3), and obtaining a mean value predicted by the system through weighted summation, wherein the mean value is as follows:

reconstructing the observation model to obtain the following nonlinear model:

wherein the content of the first and second substances,E[r(k)r^T(k)]＝R(k)

(9) left-multiplying both sides of the reconstructed observation model by S^-1(k) Due to S^-1(k)＝S^T(K) Then the following equation is obtained:

wherein the content of the first and second substances,e(k)＝S^T(k)φ(k)

then element e of line j in e (k)_j(k) Is e_j(k)＝d_j(k)-m_j(x(k))。

(10) According to the property of the correlation entropy, the cost function based on the correlation entropy is

When the function J (-) is maximal, the optimal estimate of the state can be obtained, and then the optimal estimate of x (k) is

Then, order

The above formula can be obtained:

this gives:

then, a modified covariance is obtained as

Wherein:

Φ(k)＝diag(G_Υ(e₁(k),…,G_Υ(e_n+m(k))))

(11) the actual state is not known in practice, letNamely, it isThen there are:

the modified observed covariance was:

the covariance observed by the system becomes:

then, the cross covariance matrix is:

kalman gain matrix of

(12) Finally, a state update and a covariance update are calculated

Example 1

In order to fully verify the rationality of the MCUKF method, a Zigbee indoor positioning system based on CC2530 is adopted for carrying out experiments, an unknown node and four AP points are arranged in the Zigbee indoor positioning system based on CC2530, the four AP points are randomly distributed, and the coordinates of the AP points are determined. And then, acquiring signal intensity data by using a PC (personal computer) under the static condition of an unknown node, acquiring 3min for each point, selecting 50 groups of data from the data, carrying out test analysis on the error distribution characteristics of the 4 AP signal intensities received by the unknown node, and analyzing the result of Kolmogorov-Smirnov test, wherein if the test rejects a zero hypothesis on a 15% significant level, the signal intensities from the unknown node to the 4 APs are all in non-normal distribution. The signal strength of the AP received by the unknown node is in a non-Gaussian distribution state, and the ranging errors of the unknown node and the known AP are also non-Gaussian noise. Therefore, the signal intensity distribution is superposed according to several Gaussian distributions, the signal intensity distribution is fitted through a mixed Gaussian model, and the proposed MCUKF positioning algorithm is calculated according to the distribution, as shown in FIG. 2.

As shown in figure 2, because of introducing the maximum correlation entropy, MCUKF can be known by comparing with other positioning algorithms, MCUKF reduces the requirement on the distribution characteristic of observation noise, can carry out filtering processing on observation data better, positioning accuracy is obviously improved, especially when other algorithm positioning errors are larger, the accuracy of the MCUKF positioning algorithm is obviously improved, and therefore the MCUKF positioning algorithm is stronger in robustness compared with other algorithms and stronger in adaptive environment capacity.

Example 2

In a practical and complex indoor environment, the positioning accuracy is often influenced by the multi-peak heavy-tail non-Gaussian noise, which can cause the accuracy of the traditional non-linear filtering algorithm to be reduced or even diverged, so that a non-linear robust filtering method aiming at the multi-peak heavy-tail non-Gaussian noise needs to be researchedThe maximum correlation entropy makes the filtering robust against non-gaussians. The following is a verification by a specific simulation experiment, and a suitable core width is selected by comparative analysis. Four APs are deployed in an indoor positioning area, corresponding coordinates of the four APs are A1(0,0), A2(0,10), A3(10,0) and A4(10,10), and state vectors of unknown nodes in an experiment are setThe initial true state is provided by LS (position unit m, speed unit m/s)The initial state estimation error covariance matrix is set to P (0 | 0) ═ diag ([1,0.5,1, 0.5)]) The process noise is Q-N (0,0.1), and the observed non-Gaussian noise is comprehensively analyzed according to the signal intensity of the testTo find the optimal core width, four different core widths were first selected for experimentation, with the core widths set to γ respectively₁＝5，Υ₂＝10，Υ₃＝20，Υ₄50. The results of the experiment are shown in FIGS. 3 and 4 below.

The MCUKF algorithm dynamic positioning comparison at four kernel widths as shown in fig. 3, the comparison of positioning accuracy as shown in fig. 4, and the kernel width y selected by MCUKF as shown in fig. 3 and 4₃When the indoor positioning precision is 10, the indoor positioning precision is the highest, and the performance effect is the best.

In one embodiment, an indoor positioning device is provided, the device comprising: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3;

and the state equation establishing module is used for establishing a state equation of the motion of the unknown moving target node.

And the observation model establishing module is used for establishing an observation model of the unknown moving target node motion.

The observation Sigma point set determining module is used for obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by the unscented transformation UT and corresponding weights thereof according to the state equation; and according to the predicted value in one step, calculating a new Sigma point set again through the unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set.

And the system observation predicted value determining module is used for calculating the observation predicted value of the observation Sigma point set according to the observation model and obtaining the observation predicted value of the system through weighted summation.

The Kalman gain matrix determination module is used for fusing the observation predicted value of the system according to the cost function based on the maximum correlation entropy determined by the property of the correlation entropy to obtain an observation covariance matrix and an observation cross covariance matrix for reconstructing an observation model; and determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix.

The moving target determining module is used for updating the system state quantity and the covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the updated system state quantity and the covariance matrix.

For specific definition of the indoor positioning device, reference may be made to the definition of the indoor positioning method above, and details are not described here. The modules in the indoor positioning device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, comprising: the positioning system is composed of a plurality of wireless access point AP nodes arranged indoors and an unknown moving target node, wherein the number of the wireless access point AP nodes is more than or equal to 3;

and establishing a state equation of the motion of the unknown moving target node.

And establishing an observation model of the unknown moving target node motion.

According to a state equation, obtaining a one-step predicted value and a covariance matrix of the system state quantity through a group of Sigma point sets calculated by unscented transformation UT and corresponding weights thereof; and according to the predicted value in one step, calculating a new Sigma point set again through the unscented transformation UT, substituting the new Sigma point set into the observation model, and obtaining an observation Sigma point set.

And calculating an observation predicted value of the observation Sigma point set according to the observation model, and obtaining the observation predicted value of the system through weighted summation.

According to a cost function based on the maximum correlation entropy, fusing an observation predicted value of the system to obtain an observation covariance matrix and an observation cross covariance matrix of a reconstructed observation model; and determining a Kalman gain matrix according to the observation covariance matrix and the observation cross covariance matrix.

Updating a system state quantity and a covariance matrix according to the Kalman gain matrix; and determining the position coordinates of the unknown moving target nodes according to the updated system state quantity and the covariance matrix.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features. Furthermore, the above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.