CN116482735A - High-precision positioning method for inside and outside of limited space - Google Patents

Abstract

The invention provides a high-precision positioning method inside and outside a limited space, which is based on hardware equipment such as UWB labels, IMU inertial sensors, BD receivers and the like, combines global positioning and local positioning through BP neural networks, fusion filtering algorithms and the like, and enhances the positioning stability and precision in complex environments inside and outside the limited space. Compared with the traditional positioning method in the limited space, the method can effectively inhibit the influence of factors such as non-line-of-sight and multipath effects on the positioning precision. Meanwhile, the problem of ping-pong effect in the process of switching the positioning positions can be reduced, the switching speed of the positioning method between different areas is improved, and seamless connection of positioning is realized.

Description

High-precision positioning method for inside and outside of limited space

Technical Field

The invention relates to a high-precision positioning method for the inside and outside of a limited space.

Background

The problem of high-precision positioning in a complex environment in a limited space is a technical problem all the time, and is limited by the influence of non-line-of-sight factors and the influence of multipath effects, and a single indoor positioning method has the problem that high-precision positioning cannot be realized, such as a WIFI fingerprint positioning method, an RFID positioning method and the like. In the existing positioning method, a positioning switching algorithm from the inside of the limited space to the outside of the limited space has a large ping-pong effect, and positioning delay is too large, such as a threshold switching method. However, the switching of the indoor and outdoor positioning algorithms is realized by selecting a plurality of sensing fusion methods such as machine learning, and the like, so that the high-precision positioning of the inside and outside of the limited space can be realized. The method can greatly improve the positioning precision of equipment in space-limited areas such as cities and canyons, and has wide application prospects in industries such as logistics, rescue and the like.

Disclosure of Invention

The invention provides a high-precision positioning method for the inside and outside of a limited space, which aims to solve the problems existing in the prior art.

The invention adopts the technical scheme that:

a high-precision positioning method for the inside and outside of a limited space comprises the following steps:

1) Dividing a space into a limited space inner area, a limited space outer area and a limited space inner and outer boundary area according to whether BD satellite signals can be received, arranging four UWB base stations in the limited space, binding UWB labels, IMU inertial sensors and BD receivers together and forming positioning labels;

measuring working states of BD receivers in all areas respectively, wherein the BD receivers receive longitude and latitude information of measuring positions where positioning tags are located, UWB tags receive signal values transmitted back by four UWB base stations, the signal values are used as input of BP neural network algorithm, actual measuring positions where the BD receivers are located are used as output, and therefore BP neural network training model M1 is built;

2) When the positioning tag is judged to be in the region in the limited space through the BP neural network training model M1, an IMU inertial sensor and a UWB tag are fused through an ESKF fusion filtering algorithm;

3) When the positioning tag is judged to be in the area outside the limited space through the BP neural network training model M1, an IMU inertial sensor and a BD receiver are fused for receiving through a Kalman filtering algorithm;

4) When the positioning tag is judged to be in the boundary area between the inside and outside of the limited space through the BP neural network training model M1, the working state of the BD receiver, longitude and latitude information of the measuring position where the BD receiver receives the positioning tag and signal values transmitted by four UWB base stations received by the UWB tag are used as input of a BP neural network algorithm, and the actual two-dimensional coordinate of the measuring position where the positioning tag is located is used as output, so that the BP neural network training model M2 is built; the BP neural network training model M2 is used for automatically obtaining the coordinates of the positioning points when judging that the positioning labels are in the boundary areas between the inside and outside of the limited space;

5) And (3) converting the coordinate information obtained in the steps 2) -4) into a navigation coordinate system through coordinates, transmitting the coordinate information to an upper computer through a serial port protocol, and simultaneously transmitting the coordinate information to a cloud for storage through an Internet of things module, so that storage and information query are facilitated.

Further, the coordinate system conversion relationship in step 5) is as follows: the obtained coordinate information is converted into a northeast coordinate system by a quaternion method, and a differential equation is shown as follows:

wherein:

ω＝[ω _x ω _y ω _z ] ^T ，ω _x ，ω _y ，ω _z three-axis angular velocity values output by a three-axis gyroscope in the IMU inertial sensor under a carrier coordinate system are respectively;

q＝[q ₀ q ₁ q ₂ q ₃ ] ^T is quaternion and normalized, i.e

The rotation matrix from the carrier coordinate system-b system to the navigation coordinate system-e system represented by the quaternion is:

the quaternion represents attitude information roll angle gamma, pitch angle theta and yaw angle psi of the triaxial gyroscope as follows:

θ＝asin(2(q ₀ q ₂ -q ₁ q ₃ ))

wherein atan2 () is an arctangent function, representing an angular range (-pi, pi);

by rotation variation, the measurement values of the triaxial accelerometer in the IMU inertial sensor in the carrier coordinate system-b system can be converted into the carrier coordinate system-e system:

wherein:representing acceleration in the carrier coordinate system-e system,/->The acceleration under the carrier coordinate system-b system is represented, and g represents the acceleration of the position where the positioning tag is located; for acceleration a ^e Integrating to obtain the acceleration and the position under the navigation coordinate system-e.

Further, in step 2), the ESKF fusion filtering algorithm divides the state quantities of the UWB tag and the IMU inertial sensor into three parts during use: the real state, the nominal state and the error state comprise the following specific processes:

21 Establishment of a nominal state equation;

judging and correcting the pseudo range measured by the UWB tag in an NLOS environment, measuring the motion state of the positioning tag by using an IMU inertial sensor, and updating the position, wherein the nominal state equation is as follows according to the motion equation and the quaternion attitude equation of the positioning tag:

wherein: a, a _m 、a _b The method comprises the steps that when a positioning tag moves, a measured value and a nominal value of acceleration measured by a triaxial accelerometer in an inertial sensor of the IMU are measured respectively;

ω _m 、ω _b measuring angular velocity of a triaxial gyroscope in the IMU inertial sensor and nominal value;

r is a transfer matrix from an inertial sensor of the IMU to the rotation of the inertial system;

p _k+1 a nominal position matrix at the moment k+1; p is p _k A nominal position matrix at the moment k; v _k+1 Nominal matrix for k+1 time；v _k A nominal matrix at the moment k; q _k+1 The gesture quaternion matrix at the moment k+1; q _k The matrix is a k-moment attitude quaternion matrix; g is an acceleration matrix of the position of the positioning tag;

22 Establishment of error state equation

The nominal state calculated from the information measured by the IMU inertial sensor in step 21) does not take into account noise effects, and if the nominal state is used for further calculation, the positioning error becomes larger and larger; therefore, the error of the nominal state in the step 21) needs to be modified by establishing an error state and then utilizing the error state, so that the positioning result is not affected by the error, and the process is as follows:

selecting displacement error delta _p Speed error delta _v Attitude error delta _θ Zero bias error delta omega of gyroscope _b Zero offset error delta a of accelerometer _b As an error state, it is noted that:

δx＝[δp ^T δv ^T δθ ^T δω _b ^T δa _b ^T ] ^T

then, the error equation of the IMU inertial sensor is:

wherein,,is a component of the rotational angular velocity of the earth;

at this time, error transfer matrix F _t The method comprises the following steps:

wherein,,

zero exists in use process of IMU inertial sensorPartial noise, noise matrix W and noise transfer matrix B _t The method comprises the following steps of:

W＝[α _ωx α _ωy α _ωz ω _ωx ω _ωy ω _ωz ] ^T

wherein: alpha _ωx 、α _ωy 、α _ωz Noise of the triaxial accelerometer in the IMU inertial sensor on X, Y and Z axes respectively; omega _ωx 、ω _ωy 、ω _ωy Noise of the triaxial gyroscope in the IMU inertial sensor on X, Y and Z axes respectively;

therefore, the error state equation is:

X _k+1|k ＝F _K X _k|k +B _K W _K+1 ；

in the error state equation:

X _k+1|k the k+1 moment state matrix estimated value is obtained through the k moment state matrix; f (F) _K The state transition matrix is k time; x is X _k|k Estimating a state matrix for the posterior at the moment k, wherein the state matrix is the result of the last filtering; b (B) _K For input to the state transition matrix; w (W) _k+1 An input matrix at time k+1;

23 UWB fusion IMU to correct NLOS error

After correcting the error of the nominal state in step 22), the nominal value of the positioning position can be obtained, and the pseudo-range value measured by each UWB positioning base station is compared with the value of the distance from each UWB positioning base station to the estimated position in the nominal state, thereby calculating the predicted distance d _i，pre :

d _i，pre ＝||p _t+1，pre -p _i ||

Distance measurement of UWB tagAnd the difference is made with the distance predicted value obtained by the predicted distance formula to obtain an error estimated value delta d _i :

When multipath effect occurs in NLOS, signal propagation path increases, so error estimation value Δd when multipath effect occurs _i Often positive;

the nominal position calculated by the IMU inertial sensor has errors, the error estimation value can be considered reasonable when the error estimation value is within a preset range, a threshold value gamma is set, and a Huber weight function omega (delta d) is established _i )

Wherein: gamma is a threshold value set in consideration of the nominal position error of the inertial sensor of the IMU, and when the error estimated value is smaller than the threshold value, NLOS does not occur, and the weight is 1; when the error estimation value is larger than the threshold value, the NLOS is considered to occur, the weight is set according to the error value, and the larger the error is, the smaller the weight is;

calculating the weight omega of each base station according to the above _i Reflecting the degree of NLOS of each positioning base station of UWB;

constructing a weight matrix W by the weight of each UWB positioning base station:

the updated coordinates P are obtained using a weighted least squares method:

P＝(X ^T WX) ^-1 X ^T WY

obtaining a new pseudo range according to the obtained new position coordinates, repeating the process, and finally converging to the condition that each pseudo range weight is close to 1 to obtain the pseudo range with reduced NLOS influence;

24 Error status measurement update

The IMU inertial sensor data are mixed with a large amount of noise, the error state is required to be corrected by using corrected pseudo-range information, and the basic form of an observation equation is as follows:

y＝h(x)+σ ²

wherein: y is observation data; h (x) is an observation function of the UWB positioning base station; sigma (sigma) ² Is the variance of the observed noise;

in the ESKF fusion filtering algorithm, to update the error state, it is necessary to calculate the jacobian matrix of the observation function for the error state:

the chain rule is applied, and the following steps are:

wherein:linear expansion for observation h (x);

according to the definition of the state variable in the formula, thenCan be expressed as:

wherein,,

measuring and updating through the calculated Jacobian matrix, wherein the measuring and updating comprises calculation of Kalman gain, updating of an error state covariance matrix and updating of error state measurement;

the measurement update equation in the ESKF fusion filtering algorithm is as follows:

wherein,,

K _k+1 a Kalman gain at time k+1; p (P) _k+1|k Estimating covariance a priori at time k+1; h is the transfer matrix from the state quantity to the measurement; v is a measurement noise matrix; p (P) _k+1|k+1 Estimating covariance for a posterior at time k+1; y is Y _k Is an observation matrix; x is X _k+1|k+1 Estimating a state matrix for the posterior at the moment k+1;

combining the error state after measurement and update with the nominal state to obtain the state information of the positioning target:

further, in step 3), the IMU inertial sensor and the BD receiver are used for receiving and positioning outside the limited space area, the receiving information received by the IMU inertial sensor and the BD receiver is fused by a Kalman filtering method, so as to obtain an optimal position, the fusion by the Kalman filtering method comprises a prediction step and an updating step,

the prediction steps are as follows:

X(k|k-1)＝AX(k-1|k-1)+ω

P(k|k-1)＝AP(k-1|k-1)A ^T +Q

wherein,,

x (k|k-1) is a predicted value at k time, X (k-1|k-1) is an optimal estimated value at k-1, A is an n×n state transition matrix, ω is an input noise error, P (k|k-1) is a covariance matrix at k time, P (k-1|k-1) is a correction covariance matrix in k-1 state, and Q is a process noise covariance;

the updating steps are as follows:

K(k)＝P(k|k-1)H(k) ^T (H(k)P(k|k-1)H(k) ^T +R(k)) ^-1

X(k|k)＝X(k|k-1)+K(k)(Z(k)-H(k)X(k|K-1))

P(k|k)＝(I-K(k)H)P(k|k-1)

where K is the kalman gain at time K, H is the observation equation at time K, R is the measurement error variance matrix, X (k|k) is the optimal estimate at time K, Z (K) is the measurement at time K, and P (k|k) is the correction covariance matrix at time K.

The invention has the following beneficial effects:

compared with the traditional positioning method in the limited space, the method can effectively inhibit the influence of factors such as non-line-of-sight and multipath effects on the positioning precision. Meanwhile, the problem of ping-pong effect in the process of switching the positioning positions can be reduced, the switching speed of the positioning method between different areas is improved, seamless connection of positioning is realized, and a complete positioning system inside and outside a limited space is established.

Drawings

FIG. 1 is a flow chart of a method according to an embodiment of the invention

Detailed Description

The invention is further described below with reference to the accompanying drawings.

As shown in FIG. 1, the method for positioning the inside and outside of the limited space with high precision comprises the following steps:

1) Selecting a limited space with weak GNSS signals, dividing the space into a limited space inner area, a limited space outer area and a limited space inner and outer interface area according to whether BD satellite signals can be received, arranging four UWB base stations in the limited space, binding UWB labels, IMU inertial sensors and BD receivers together and forming positioning labels;

the working states of BD receivers are measured in all areas respectively, the BD receivers receive longitude and latitude information of measuring positions where positioning tags are located, UWB tags receive signal values transmitted back by four UWB base stations, the signal values are used as input of BP neural network algorithm, actual measuring positions where the BD receivers are located are used as output, and therefore BP neural network training model M1 is built.

2) When the positioning tag is judged to be in the region in the limited space through the BP neural network training model M1, the IMU inertial sensor and the UWB tag are fused through the ESKF fusion filtering algorithm, and the UWB tag can achieve higher positioning accuracy in the closed space, but is also susceptible to factors such as non-line-of-sight and multipath effects. To solve this problem, the present invention uses the method of ESKF fusion filtering fusion IMU and UWB to improve positioning accuracy in a confined space. The main idea is as follows: after the measurement data of the IMU inertial sensor and the UWB tag are obtained, the measurement data of the IMU inertial sensor is subjected to position prediction in an ESKF prediction link, and a prediction error state and a nominal state are obtained. And comparing the pseudo range measured by each UWB base station with the distance from each base station to the estimated position in the nominal state, calculating the weight of each base station, and carrying out iterative re-weighting least square processing on the pseudo range by taking the weight as an initial value to obtain a corrected pseudo range. And finally, correcting the error state by the corrected pseudo range to obtain the final calculated position.

The specific algorithm comprises the following four steps: (1) establishment of a nominal state equation; (2) establishment of an error state equation; (3) The UWB fusion IMU corrects NLOS error (4) error status measurement update.

21 Establishment of a nominal state equation;

judging and correcting the pseudo range measured by the UWB tag in an NLOS environment, measuring the motion state of the positioning tag by using an IMU inertial sensor, and updating the position, wherein the nominal state equation is as follows according to the motion equation and the quaternion attitude equation of the positioning tag:

wherein: a, a _m 、a _b The method comprises the steps that when a positioning tag moves, a measured value and a nominal value of acceleration measured by a triaxial accelerometer in an inertial sensor of the IMU are measured respectively;

ω _m 、ω _b measuring angular velocity of a triaxial gyroscope in the IMU inertial sensor and nominal value;

r is a transfer matrix from an inertial sensor of the IMU to the rotation of the inertial system;

p _k+1 a nominal position matrix at the moment k+1; p is p _k A nominal position matrix at the moment k; v _k+1 A nominal matrix at the moment k+1; v _k A nominal matrix at the moment k; q _k+1 The gesture quaternion matrix at the moment k+1; q _k The matrix is a k-moment attitude quaternion matrix; g is an acceleration matrix of the position of the positioning tag;

22 Establishment of error state equation

The nominal state calculated from the information measured by the IMU inertial sensor in step 21) does not take into account noise effects, and if the nominal state is used for further calculation, the positioning error becomes larger and larger; therefore, the error of the nominal state in the step 21) needs to be modified by establishing an error state and then utilizing the error state, so that the positioning result is not affected by the error, and the process is as follows:

selecting displacement error delta _p Speed error delta _v Attitude error delta _θ Zero bias error delta omega of gyroscope _b Zero offset error delta a of accelerometer _b As an error state, it is noted that:

δx＝[δp ^T δv ^T δθ ^T δω _b ^T δa _b ^T ] ^T

then, the error equation of the IMU inertial sensor is:

wherein,,is a component of the rotational angular velocity of the earth;

at this time, error transfer matrix F _t The method comprises the following steps:

wherein,,

zero bias noise exists in the use process of the IMU inertial sensor, and the noise matrix W and the noise transfer matrix B _t The method comprises the following steps of:

W＝[α _ωx α _ωy α _ωz ω _ωx ω _ωy ω _ωz ] ^T

wherein: alpha _ωx 、α _ωy 、α _ωz Noise of the triaxial accelerometer in the IMU inertial sensor on X, Y and Z axes respectively; omega _ωx 、ω _ωy 、ω _ωy Noise of the triaxial gyroscope in the IMU inertial sensor on X, Y and Z axes respectively;

therefore, the error state equation is:

X _k+1|k ＝F _K X _k|k +B _K W _K+1 ；

in the error state equation:

X _k+1|k the k+1 moment state matrix estimated value is obtained through the k moment state matrix; f (F) _K The state transition matrix is k time; k (K) _k|k Estimating a state matrix for the posterior at the moment k, wherein the state matrix is the result of the last filtering; b (B) _K For input to the state transition matrix; w (W) _k+1 An input matrix at time k+1;

23 UWB fusion IMU to correct NLOS error

After correcting the error of the nominal state in step 22), the nominal value of the positioning position can be obtained, and the pseudo-range value measured by each UWB positioning base station is compared with the value of the distance from each UWB positioning base station to the estimated position in the nominal state, thereby calculating the predicted distance d _i，pre :

d _i，pre ＝||p _t+1，pre -p _i ||

Distance measurement of UWB tagValue ofAnd the difference is made with the distance predicted value obtained by the predicted distance formula to obtain an error estimated value delta d _i :

When multipath effect occurs in NLOS, signal propagation path increases, so error estimation value Δd when multipath effect occurs _i Often positive;

the nominal position calculated by the IMU inertial sensor has errors, the error estimation value can be considered reasonable when the error estimation value is within a preset range, a threshold value gamma is set, and a Huber weight function omega (delta d) is established _i )

Wherein: gamma is a threshold value set in consideration of the nominal position error of the inertial sensor of the IMU, and when the error estimated value is smaller than the threshold value, NLOS does not occur, and the weight is 1; when the error estimation value is larger than the threshold value, the NLOS is considered to occur, the weight is set according to the error value, and the larger the error is, the smaller the weight is;

calculating the weight omega of each base station according to the above _i Reflecting the degree of NLOS of each positioning base station of UWB;

constructing a weight matrix W by the weight of each UWB positioning base station:

the updated coordinates P are obtained using a weighted least squares method:

P＝(X ^T WX) ^-1 X ^T WY

obtaining a new pseudo range according to the obtained new position coordinates, repeating the process, and finally converging to the condition that each pseudo range weight is close to 1 to obtain the pseudo range with reduced NLOS influence;

24 Error status measurement update

The IMU inertial sensor data are mixed with a large amount of noise, the error state is required to be corrected by using corrected pseudo-range information, and the basic form of an observation equation is as follows:

y＝h(x)+σ ² wherein: y is observation data; h (x) is an observation function of the UWB positioning base station; sigma (sigma) ² Is the variance of the observed noise;

in the ESKF fusion filtering algorithm, to update the error state, it is necessary to calculate the jacobian matrix of the observation function for the error state:

the chain rule is applied, and the following steps are:

wherein:linear expansion for observation h (x);

according to the definition of the state variable in the formula, thenCan be expressed as:

wherein,,

measuring and updating through the calculated Jacobian matrix, wherein the measuring and updating comprises calculation of Kalman gain, updating of an error state covariance matrix and updating of error state measurement;

the measurement update equation in the ESKF fusion filtering algorithm is as follows:

wherein,,

K _k+1 a Kalman gain at time k+1; p (P) _k+1|k Estimating covariance a priori at time k+1; h is the transfer matrix from the state quantity to the measurement; v is a measurement noise matrix; p (P) _k+1|k+1 Estimating covariance for a posterior at time k+1; y is Y _k Is an observation matrix; x is X _k+1|k+1 Estimating a state matrix for the posterior at the moment k+1;

combining the error state after measurement and update with the nominal state to obtain the state information of the positioning target:

3) When the positioning tag is judged to be in the area outside the limited space through the BP neural network training model M1, the GNSS signals are good at the moment, and when the positioning tag is judged to be outside the limited space through the model M1, the GNSS signals are good at the moment, so that the positioning is performed by using a fusion algorithm of the IMU inertial sensor and the BD receiver at the moment. And the received values of the IMU inertial sensor and the BD receiver are fused through a Kalman filtering algorithm, so that the ping-pong effect generated in the transition process is relieved. The specific derivation process of Kalman is as follows:

the prediction steps are as follows:

X(k|k-1)＝AX(k-1|k-1)+ω

P(k|k-1)＝AP(k-1|k-1)A ^T +Q

wherein,,

x (k|k-1) is a predicted value at k time, X (k-1|k-1) is an optimal estimated value at k-1, A is an n×n state transition matrix, ω is an input noise error, P (k|k-1) is a covariance matrix at k time, P (k-1|k-1) is a correction covariance matrix in k-1 state, and Q is a process noise covariance;

the updating steps are as follows:

K(k)＝P(k|k-1)H(k)T(H(k)P(k|k-1)H(k) ^T +R(k)) ^-1

X(k|k)＝X(k|k-1)+K(k)(Z(k)-H(k)X(k|k-1))

P(k|k)＝(I-K(k)H)P(k|k-1)

where K is the kalman gain at time K, H is the observation equation at time K, R is the measurement error variance matrix, X (k|k) is the optimal estimate at time K, Z (K) is the measurement at time K, and P (k|k) is the correction covariance matrix at time K.

4) When the positioning tag is judged to be in the boundary area between the inside and outside of the limited space through the BP neural network training model M1, the working state of the BD receiver, longitude and latitude information of the measuring position where the BD receiver receives the positioning tag and signal values transmitted by four UWB base stations received by the UWB tag are used as input of a BP neural network algorithm, and the actual two-dimensional coordinate of the measuring position where the positioning tag is located is used as output, so that the BP neural network training model M2 is built; the BP neural network training model M2 is used for automatically obtaining the coordinates of the positioning points when judging that the positioning labels are in the boundary areas between the inside and outside of the limited space;

5) And (3) converting the coordinate information obtained in the steps 2) -4) into a navigation coordinate system through coordinates, transmitting the coordinate information to an upper computer through a serial port protocol, and simultaneously transmitting the coordinate information to a cloud for storage through an Internet of things module, so that storage and information query are facilitated.

The invention converts the obtained coordinate information into a northeast coordinate system by a quaternion method, and the differential equation is as follows:

/>

wherein:

ω＝[ω _x ω _y ω _z ] ^T ，ω _x ，ω _y ，ω _z three-axis angular velocity values output by a three-axis gyroscope in the IMU inertial sensor under a carrier coordinate system are respectively;

q＝[q ₀ q ₁ q ₂ q ₃ ] ^T is quaternion and normalized, i.e

The rotation matrix from the carrier coordinate system-b system to the navigation coordinate system-e system represented by the quaternion is:

the quaternion represents attitude information roll angle gamma, pitch angle theta and yaw angle psi of the triaxial gyroscope as follows:

θ＝asin(2(q ₀ q ₂ -q ₁ q ₃ ))

wherein atan2 () is an arctangent function, representing an angular range (-pi, pi);

by rotation variation, the measurement values of the triaxial accelerometer in the IMU inertial sensor in the carrier coordinate system-b system can be converted into the carrier coordinate system-e system:

wherein:representing acceleration in the carrier coordinate system-e system,/->The acceleration under the carrier coordinate system-b system is represented, and g represents the acceleration of the position where the positioning tag is located; for acceleration a ^e Integrating to obtain the acceleration and the position under the navigation coordinate system-e.

The foregoing is merely a preferred embodiment of the invention, and it should be noted that modifications could be made by those skilled in the art without departing from the principles of the invention, which modifications would also be considered to be within the scope of the invention.

Claims (4)

1. A high-precision positioning method for the inside and outside of a limited space is characterized in that: the method comprises the following steps:

1) Dividing a space into a limited space inner area, a limited space outer area and a limited space inner and outer boundary area according to whether BD satellite signals can be received, arranging four UWB base stations in the limited space, binding UWB labels, IMU inertial sensors and BD receivers together and forming positioning labels;

measuring working states of BD receivers in all areas respectively, wherein the BD receivers receive longitude and latitude information of measuring positions where positioning tags are located, UWB tags receive signal values transmitted back by four UWB base stations, the signal values are used as input of BP neural network algorithm, actual measuring positions where the BD receivers are located are used as output, and therefore BP neural network training model M1 is built;

2) When the positioning tag is judged to be in the region in the limited space through the BP neural network training model M1, an IMU inertial sensor and a UWB tag are fused through an ESKF fusion filtering algorithm;

3) When the positioning tag is judged to be in the area outside the limited space through the BP neural network training model M1, an IMU inertial sensor and a BD receiver are fused for receiving through a Kalman filtering algorithm;

4) When the positioning tag is judged to be in the boundary area between the inside and outside of the limited space through the BP neural network training model M1, the working state of the BD receiver, longitude and latitude information of the measuring position where the BD receiver receives the positioning tag and signal values transmitted by four UWB base stations received by the UWB tag are used as input of a BP neural network algorithm, and the actual two-dimensional coordinate of the measuring position where the positioning tag is located is used as output, so that the BP neural network training model M2 is built; the BP neural network training model M2 is used for automatically obtaining the coordinates of the positioning points when judging that the positioning labels are in the boundary areas between the inside and outside of the limited space;

5) And (3) converting the coordinate information obtained in the steps 2) -4) into a navigation coordinate system through coordinates, transmitting the coordinate information to an upper computer through a serial port protocol, and simultaneously transmitting the coordinate information to a cloud for storage through an Internet of things module, so that storage and information query are facilitated.

2. The high-precision positioning method for the inside and outside of a limited space according to claim 1, wherein: the coordinate system conversion relation in the step 5) is as follows: the obtained coordinate information is converted into a northeast coordinate system by a quaternion method, and a differential equation is shown as follows:

wherein:

ω＝[ω _x ω _y ω _z ] ^T ，ω _x ，ω _y ，ω _z three-axis angular velocity values output by a three-axis gyroscope in the IMU inertial sensor under a carrier coordinate system are respectively;

q＝[q ₀ q ₁ q ₂ q ₃ ] ^T is quaternion and normalized, i.e

The rotation matrix from the carrier coordinate system-b system to the navigation coordinate system-e system represented by the quaternion is:

the quaternion represents attitude information roll angle gamma, pitch angle theta and yaw angle psi of the triaxial gyroscope as follows:

θ＝asin(2(q ₀ q ₂ -q ₁ q ₃ ))

wherein atan2 () is an arctangent function, representing an angular range (-pi, pi);

by rotation variation, the measurement values of the triaxial accelerometer in the IMU inertial sensor in the carrier coordinate system-b system can be converted into the carrier coordinate system-e system:

wherein:representing acceleration in the carrier coordinate system-e system,/->The acceleration under the carrier coordinate system-b system is represented, and g represents the acceleration of the position where the positioning tag is located; for acceleration a ^e Integrating to obtain the acceleration and the position under the navigation coordinate system-e.

3. The high-precision positioning method for the inside and outside of the limited space as claimed in claim 2, wherein: in the step 2), the ESKF fusion filtering algorithm divides the state quantity of the UWB tag and the IMU inertial sensor into three parts in the use process: the real state, the nominal state and the error state comprise the following specific processes:

21 Establishment of a nominal state equation;

judging and correcting the pseudo range measured by the UWB tag in an NLOS environment, measuring the motion state of the positioning tag by using an IMU inertial sensor, and updating the position, wherein the nominal state equation is as follows according to the motion equation and the quaternion attitude equation of the positioning tag:

wherein: a, a _m 、a _b The method comprises the steps that when a positioning tag moves, a measured value and a nominal value of acceleration measured by a triaxial accelerometer in an inertial sensor of the IMU are measured respectively;

ω _m 、ω _b measuring angular velocity of a triaxial gyroscope in the IMU inertial sensor and nominal value;

r is a transfer matrix from an inertial sensor of the IMU to the rotation of the inertial system;

p _k+1 a nominal position matrix at the moment k+1; p is p _k A nominal position matrix at the moment k; v _k+1 A nominal matrix at the moment k+1; v _k A nominal matrix at the moment k; q _k+1 The gesture quaternion matrix at the moment k+1; q _k The matrix is a k-moment attitude quaternion matrix; g is an acceleration matrix of the position of the positioning tag;

22 Establishment of error state equation

The nominal state calculated from the information measured by the IMU inertial sensor in step 21) does not take into account noise effects, and if the nominal state is used for further calculation, the positioning error becomes larger and larger; therefore, the error of the nominal state in the step 21) needs to be modified by establishing an error state and then utilizing the error state, so that the positioning result is not affected by the error, and the process is as follows:

selecting displacement error delta _p Speed error delta _v Attitude error delta _θ Zero bias error delta omega of gyroscope _b Zero offset error delta a of accelerometer _b As an error state, it is noted that:

δx＝[δp ^T δv ^T δθ ^T δω _b ^T δa _b ^T ] ^T

then, the error equation of the IMU inertial sensor is:

wherein,,is a component of the rotational angular velocity of the earth;

at this time, error transfer matrix F _t The method comprises the following steps:

wherein,,

zero bias noise exists in the use process of the IMU inertial sensor, and the noise matrix W and the noise transfer matrix B _t The method comprises the following steps of:

W＝[α _ωx α _ωy α _ωz ω _ωx ω _ωy ω _ωz ] ^T

wherein: alpha _ωx 、α _ωy 、α _ωz Noise of the triaxial accelerometer in the IMU inertial sensor on X, Y and Z axes respectively; omega _ωx 、ω _ωy 、ω _ωy Noise of the triaxial gyroscope in the IMU inertial sensor on X, Y and Z axes respectively;

therefore, the error state equation is:

X _k+1|k ＝F _K X _k|k +B _K W _K+1 ；

in the error state equation:

X _k+1|k the k+1 moment state matrix estimated value is obtained through the k moment state matrix; f (F) _K The state transition matrix is k time; x is X _k|k Estimating a state matrix for the posterior at the moment k, wherein the state matrix is the result of the last filtering; b (B) _K For input to the state transition matrix; w (W) _k+1 An input matrix at time k+1;

23 UWB fusion IMU to correct NLOS error

After correcting the error of the nominal state in step 22), the nominal value of the positioning position can be obtained, and the pseudo-range value measured by each UWB positioning base station is compared with the value of the distance from each UWB positioning base station to the estimated position in the nominal state, thereby calculating the predicted distance d _i，pre ：

d _i，pre ＝||p _t+1，pre -p _i ||

Distance measurement d of UWB tag _i And the difference is made with the distance predicted value obtained by the predicted distance formula to obtain an error estimated value delta d _i ：

When multipath effect occurs in NLOS, signal propagation path increases, so error estimation value Δd when multipath effect occurs _i Often positive;

the nominal position calculated by the IMU inertial sensor has errors, the error estimation value can be considered reasonable when the error estimation value is within a preset range, a threshold value gamma is set, and a Huber weight function omega (delta d) is established _i )

Wherein: gamma is a threshold value set in consideration of the nominal position error of the inertial sensor of the IMU, and when the error estimated value is smaller than the threshold value, NLOS does not occur, and the weight is 1; when the error estimation value is larger than the threshold value, the NLOS is considered to occur, the weight is set according to the error value, and the larger the error is, the smaller the weight is;

calculating the weight omega of each base station according to the above _i Reflecting the degree of NLOS of each positioning base station of UWB;

constructing a weight matrix W by the weight of each UWB positioning base station:

the updated coordinates P are obtained using a weighted least squares method:

P＝(X ^T WX) ^-1 X ^T WY

obtaining a new pseudo range according to the obtained new position coordinates, repeating the process, and finally converging to the condition that each pseudo range weight is close to 1 to obtain the pseudo range with reduced NLOS influence;

24 Error status measurement update

The IMU inertial sensor data are mixed with a large amount of noise, the error state is required to be corrected by using corrected pseudo-range information, and the basic form of an observation equation is as follows:

y＝h(x)+σ ²

wherein: y is observation data; h (x) is an observation function of the UWB positioning base station; sigma (sigma) ² Is the variance of the observed noise;

in the ESKF fusion filtering algorithm, to update the error state, it is necessary to calculate the jacobian matrix of the observation function for the error state:

the chain rule is applied, and the following steps are:

wherein:linear expansion for observation h (x);

according to the definition of the state variable in the formula, thenCan be expressed as:

wherein,,

measuring and updating through the calculated Jacobian matrix, wherein the measuring and updating comprises calculation of Kalman gain, updating of an error state covariance matrix and updating of error state measurement;

the measurement update equation in the ESKF fusion filtering algorithm is as follows:

wherein,,

K _k+1 a Kalman gain at time k+1; p (P) _k+1|k Estimating covariance a priori at time k+1; h is the transfer matrix from the state quantity to the measurement; v is a measurement noise matrix; p (P) _k+1|k+1 Estimating covariance for a posterior at time k+1; y is Y _k Is an observation matrix; x is X _k+1|k+1 Estimating a state matrix for the posterior at the moment k+1;

combining the error state after measurement and update with the nominal state to obtain the state information of the positioning target:

4. a limited space inside and outside high precision positioning method as claimed in claim 3, characterized in that: in the step 3), the IMU inertial sensor and the BD receiver are used for receiving and positioning at the same time outside the limited space area, the receiving information received by the IMU inertial sensor and the BD receiver is fused by a Kalman filtering method, so as to obtain an optimal position, the fusion by the Kalman filtering method comprises a prediction step and an updating step,

the prediction steps are as follows:

X(k|k-1)＝AX(k-1|k-1)+ω

P(k|k-1)＝AP(k-1|k-1)A ^T +Q

wherein,,

x (k|k-1) is a predicted value at k time, X (k-1|k-1) is an optimal estimated value at k-1, A is an n×n state transition matrix, ω is an input noise error, P (k|k-1) is a covariance matrix at k time, P (k-1|k-1) is a correction covariance matrix in k-1 state, and Q is a process noise covariance;

the updating steps are as follows:

K(k)＝P(k|k-1)H(k) ^T (H(k)P(k|k-1)H(k) ^T +R(k)) ^-1

X(k|k)＝X(k|k-1)+K(k)(Z(k)-H(k)X(k|k-1))

P(k|k)＝(I-K(k)H)P(k|k-1)

where K is the kalman gain at time K, H is the observation equation at time K, R is the measurement error variance matrix, X (k|k) is the optimal estimate at time K, Z (K) is the measurement at time K, and P (k|k) is the correction covariance matrix at time K.