CN112180323A

CN112180323A - Wi-Fi-based TOA and AOA indoor combined positioning algorithm research

Info

Publication number: CN112180323A
Application number: CN202010932469.4A
Authority: CN
Inventors: 何维; 宋雅倩; 田增山; 岳紫颖
Original assignee: Chongqing University of Post and Telecommunications
Current assignee: Chongqing University of Post and Telecommunications
Priority date: 2020-09-08
Filing date: 2020-09-08
Publication date: 2021-01-05

Abstract

The invention provides an indoor ToA (time of arrival) and AoA (angle of arrival) combined positioning algorithm suitable for NLOS (Non Line of Sight, NLOS) environment, which can effectively utilize the ToA and the AoA to position a target. Firstly, an NLOS error in ToA is more accurately identified by using Kalman filtering of a measurement value discarding method, then an improved Threshold comparative weighted algorithm (TCW) is constructed according to the ToA and the AoA obtained by measurement, the initial position of a target node is calculated by improving the TCW algorithm, and an experimental result shows that the positioning precision reaches 90% and the error is within 1.5 m.

Description

Wi-Fi-based TOA and AOA indoor combined positioning algorithm research

Technical Field

The invention belongs to an indoor positioning technology, and provides a ToA (time of arrival) and AoA (angle of arrival) indoor combined positioning algorithm model suitable for Wi-Fi, aiming at the problem of low positioning accuracy caused by Non-Line of Sight (NLOS) propagation of wireless signals in an indoor complex environment.

Background

With the development and popularization of wireless communication technology and network technology, people apply a great deal of location information in life, for example, to provide accurate synchronous clock sources for telecommunication base stations, television transmitting stations, and the like; the method is widely applied to positioning systems such as aviation transportation, electronic maps, navigation of agricultural implements, automatic driving, high-precision land leveling navigation and the like; a large amount of GPS equipment is adopted for engineering measurement in the construction of roads, bridges and tunnels; and the application in the field of field exploration and urban planning. The outdoor positioning system basically meets some positioning requirements of people, but people also have great requirements on indoor position information, for example, people, equipment and materials can be positioned in real time in factories, construction sites and other places; navigating the position of the shop in a shopping mall; in a prison, positioning a prisoner in real time; in the aspect of security, indoor positioning technology is mainly applied to the tracking of suspicious objects. However, due to the complexity of the indoor environment and the severe attenuation of signals, some indoor positioning problems are difficult to solve by the outdoor positioning system. Therefore, indoor positioning technology becomes a research hotspot.

Currently, the indoor positioning technologies include ultra-wideband positioning technology, fingerprint positioning, Radio Frequency Identification (RFID) positioning, bluetooth positioning, and Wi-Fi positioning. However, ultra-wideband positioning equipment is expensive, a large amount of measurement work needs to be performed in the early stage based on fingerprint positioning, radio frequency identification has no communication capability, the anti-interference capability is poor, the stability of a bluetooth system is slightly poor, interference by noise signals is large, and the price of devices and equipment is expensive. Wi-Fi is provided in most indoor environments, and the hardware cost is low, so that the positioning technology based on Wi-Fi can be widely applied.

The positioning parameters commonly used for Wi-Fi are mainly: received Signal Strength (RSS), Angle of Arrival (AoA), and Time of Arrival (ToA). Positioning based on signal strength calculates the signal strength between the transmitter and the receiver through a path propagation model, then converts the signal strength into a distance, and finally calculates the position of a target point through trilateration. The time of flight of the receiver and transmitter signals is estimated based on the time of arrival positioning, and then positioning is performed by using a trilateration algorithm, and the positioning accuracy is limited by the time estimation accuracy. The positioning based on the arrival angle needs to estimate the arrival angle of the signal to a receiver, and then the target is positioned by utilizing a trilateral positioning method. The classical approach to solving the angle of arrival is the MUSIC algorithm, which requires that the number of physical antennas of the system must be greater than the number of multipath signals.

The combined positioning algorithm based on Threshold Comparative Weighted (TCW) -Taylor series expansion mentioned in the research on indoor combined positioning algorithm based on ToA and AoA is improved. Firstly, NLOS errors in ToA are more accurately identified through measured value discarding method Kalman filtering, then initial position calculation of a target node is carried out through a proposed improved TCW method according to ToA and AoA obtained through measurement, simulation results show that positioning accuracy of an algorithm of the proposed improved threshold weighting method (improved TCW) is improved, and the initial positioning result reaches 90% and the error is within 1.5 m.

Disclosure of Invention

The invention aims to provide an indoor ToA and AoA combined positioning algorithm suitable for an NLOS environment, which can effectively utilize ToA and AoA information to position a target.

The invention relates to a method for constructing an indoor positioning model, which comprises the following steps:

step one, constructing an exponential model of NLOS errors;

step two, eliminating NLOS errors in the ToA by using Kalman filtering (Kalman Filter, KF) of a measured value discarding method, wherein a basic equation of the KF is a recursion form, and continuously predicting and correcting by using a previous state;

and step three, constructing an improved threshold comparison weighting algorithm model.

Advantageous effects

The invention aims to provide an indoor ToA and AoA combined positioning algorithm suitable for NLOS environment, which can effectively utilize ToA and AoA information to position a target and has the following advantages:

1. the NLOS error in the ToA is estimated more accurately by adopting Kalman filtering of a measurement value discarding method;

2. the measurement error of the AoA is allowed to be larger;

3. the positioning accuracy of the improved threshold weighting method (improved TCW) algorithm is improved, and the initial positioning result reaches 90% with the error within 1.5 m.

Drawings

FIG. 1 is a flow chart of the present invention.

Fig. 2 is a graph of the difference between the estimated NLOS error and the true NLOS error.

FIG. 3 is a schematic diagram of an improved TCW process.

FIG. 4 is a graph of a probability accumulation of errors in TCW versus errors in the TCW method.

Detailed description of the preferred embodiments

The invention is described in further detail below with reference to the accompanying drawings:

FIG. 1 is a flow chart of the present invention, comprising the following steps:

step one, constructing an NLOS error model, wherein NLOS errors come from indoor complex environments and obstacles, and when the signals meet the obstacles, the NLOS errors are reflected, so that the signals reaching a receiver are synthesized by a plurality of error signals, and further, the ToA and AoA measured values have large errors. To better study NLOS errors, a T1PI model proposed by a localization technology research group of Ericsson, which is a channel model commonly used in simulation evaluation based on time localization technology:

wherein D (tau) is a non-line-of-sight transmission delay, tau_rmsThe delay speed is a variable which follows a lognormal distribution and can be defined as:

τ_rms＝T₁dξ (2)

in the formula T₁τ when d equals 1km_rmsThe unit of the median value of (d) and the distance from the receiver to the transmitter is kilometers, the unit is an exponential factor, the value is 0.5-1.0, xi is a logarithmic random variable, lg xi is the mean value of 0, the standard deviation is 2-6dB, and the following table 1 is a specific model parameter under different environments.

TABLE 1 model parameters for different channel environments

Environment(s)	T₁/us	ε	ξ/dB
				Downtown area	1.0	0.5	4
General urban area	0.4	0.5	4
				Suburb	0.3	0.5	4
Suburb	0.1	0.5	4

Step two, using a measurement value discarding method Kalman filtering to eliminate NLOS errors in ToA, wherein Kalman filtering is to optimally estimate the system state to ensure that the optimal estimation value of the system has the minimum mean square error, the main steps are divided into two steps, a prediction equation and a correction equation are constructed, two equations are used for respectively representing the transfer process of an unknown state and the relation between the input and the output of a measurement system, so that the state value at a certain moment is related to the measurement values at the current moment and the previous moment, and the state equation and the measurement equation are shown in formulas (3) and (4):

X(k+1)＝AX(k)+w(k) (3)

Z(k)＝HX(k)+v(k) (4)

in order to mitigate and eliminate NLOS error, ToA and its first derivative and NLOS error are used as the state vector to be estimated, and the relation between the state vector and the measurement vector is shown

H＝[10α]，v(k)＝n_m(k) In the formula, tau is ToA value, b (k) and n_m(k) NLOS error and measurement error, respectively, Δ is the sampling interval, z (k) is the measured value, and α is the trial value. The process of Kalman filtering is as follows (5) - (9):

P_k ^-＝AP_k-1A^T+Q (6)

P_k＝(I-K_kC)P_K ^- (9)

wherein,

and

respectively generation by generationRepresenting predicted and estimated values of the k-th time state variable, P_k ^-And P_kRespectively representing the predicted and estimated mean square error matrices at the kth time, Q and R are respectively the process noise covariance and the measured noise vector covariance, K_kIs the gain of kalman filtering at time k.

In practice, the solution of the kalman filter is calculated recursively, each updated estimate of the state is calculated from the previous estimate and from new input data, and when the previous ToA is severely affected by the NLOS error, the subsequent measurement value will have a very large error. Another problem is that the delay produced by NLOS can only be larger than zero. To solve this problem, the judgment in (8) to (9) is improved

When in use

Greater than 10ns or less than 0, or when

Is greater than 1.2 of the total weight of the rubber,

is used as

Instead of this. As shown in fig. 2, the NLOS error in the TOA estimated by kalman filtering using the measurement rejection method is better than the NLOS error estimated by the ordinary kalman filtering.

Step three, adopting the proposed improved threshold comparison weighting method, as shown in fig. 3, the specific method is as follows:

a) firstly, find out the point D where the three circles intersect_i(x_i,y_i) I is 1,2,3, 4, 5, 6, and find the common chord segment L1 where each two circles intersect_i＝k1_ix+b1_i,i＝1,2,3；

b) Making three lines of AoA, taking the circle center as one point, measuring the value of AoA as a slope, and setting the error obedient mean value of the slope as 0The difference is 1 DEG, L2_i＝k2_ix+b2_i,i＝1,2,3；

c) Find L1_iAnd L2_iIntersecting points and screening out P in a common region where the three circles intersect_i(x_i,y_i),i＝1,…,N；

d) Calculating any two intersection points P_mAnd P_nA distance d between_mn；

e) All distances d_mnIs taken as the threshold value D_thr；

f) Then all possible intersections P_iInitial weight of (I)_kIs 0, i.e. I_k0; comparison d_mnAnd D_thrIf d is_mn<D_thrThen, I_m＝I_m+1,I_n＝I_n+1,1≤m,n≤N；

g) Initial point position calculation formula (x)_t,y_t)，

Claims

1. The invention aims to provide an indoor ToA and AoA combined positioning algorithm suitable for NLOS environment, which can effectively utilize ToA and AoA to position a target and is characterized in that:

a) constructing an exponential model of NLOS errors;

b) eliminating NLOS error in ToA by adopting Kalman filtering of a measurement value discarding method;

c) and performing initial positioning by adopting a method of improving threshold comparison weighting.

2. Constructing an NLOS error model according to claim 1, where the NLOS error originates from a complex indoor environment and an obstacle, and is reflected when the signal encounters the obstacle, resulting in a signal arriving at the receiver being synthesized from multiple error signals and causing a large error between ToA and AoA measurements, and in order to better study the NLOS error, the T1PI model proposed by Ericsson's positioning technology research group is a channel model commonly used for simulation evaluation based on time-based positioning technology:

τ_rms＝T₁dξ (2)

in the formula T₁τ when d equals 1km_rmsD is the distance from the receiver to the transmitter, in kilometers; the index factor is 0.5-1.0, xi is a logarithmic random variable, lg xi is an average value of 0, the standard deviation is 2-6dB, and the following table 1 is a specific model parameter under different environments.

TABLE 1 model parameters for different channel environments

Environment(s) T₁/us ε ξ Downtown area 1.0 0.5 4 General urban area 0.4 0.5 4 Suburb 0.3 0.5 4 Suburb 0.1 0.5 4

3. The method for eliminating NLOS error in ToA by using Kalman filtering of measurement discarding method according to claim 1, wherein Kalman filtering is to optimally estimate the system state to make the optimal estimation value of the system have minimum mean square error, the main steps are divided into two steps, a prediction equation and a correction equation are constructed, the two equations are used to respectively express the transfer process of unknown state and the relation between input and output of the measurement system, thereby connecting the state value at a certain moment with the measurement value at the current and previous moments, and the state equation and the measurement equation are as shown in formulas (3) and (4):

X(k+1)＝AX(k)+w(k) (3)

Z(k)＝HX(k)+v(k) (4)

H＝[1 0 α]，v(k)＝n_m(k) In that respect In which τ is the ToA value, b (k) and n_m(k) Respectively NLOS error and measurement error, Delta is sampling interval, Z (k)Is a measured value, alpha is a test value, and the process of Kalman filtering is as in formulas (5) - (9)

P_k ^-＝AP_k-1A^T+Q (6)

P_k＝(I-K_kC)P_K ^- (9)

Wherein,

and

respectively representing the predicted and estimated values, P, of the k-th time state variable_k ^-And P_kRespectively representing the predicted and estimated mean square error matrices at the kth time, Q and R are respectively the process noise covariance and the measured noise vector covariance, K_kThe gain of Kalman filtering at the kth moment;

in practice, the solution of the kalman filter is calculated recursively, each updated estimation of the state is calculated from the previous estimation and new input data, when the previous ToA is seriously affected by the NLOS error, the subsequent measurement value will generate a very large error, and another problem is that the delay generated by the NLOS can only be greater than zero, and for this problem, improvements in (8) to (9) are made to determine the delay

When in use

Greater than 10ns or less than 0, or when

Is greater than 1.2 of the total weight of the rubber,

is used as

Instead, as shown in fig. 2, the NLOS error in the ToA estimated by kalman filtering using the measurement value discarding method is better than the NLOS error estimated by the ordinary kalman filtering.

4. The proposed improved threshold comparison weighting method is adopted according to claim 1, as shown in fig. 3, the specific method is as follows:

b) Making three lines of AoA, taking the circle center as one point, measuring the value of AoA as a slope, the error obedient mean value of the slope is 0, the variance is 1 degree, and L2_i＝k2_ix+b2_i,i＝1,2,3；

d) Calculating any two intersection points P_mAnd P_nA distance d between_mn. All distances d_mnIs taken as the threshold value D_thr；

e) Let all possible intersections P_iInitial weight of (I)_kIs 0, i.e. I_k0; comparison d_mnAnd D_thrIf d is_mn<D_thrThen, I_m＝I_m+1,I_n＝I_n+1,1≤m,n≤N；

f) Initial point position calculation formula (x)_t,y_t)，