An interactive multi-model indoor positioning method based on adaptive unscented kalman filter

Abstract

The application provides an interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering, and is realized based on a positioning calculation terminal device; the method comprises the following steps: S1, obtaining the time of flight between a beacon node and a target to be positioned based on a ranging sensor module, and then calculating the distance information between the two, and providing the distance information to a master module through a routing node; S2, the master module sends the distance information from the ranging sensor module to a server for storage and processing through a coordinator, and finally calculates the position coordinates in the host computer through an IMM-AUKF algorithm; S3, the display module stores and displays the position coordinates and the moving track of the target to be positioned calculated in the master module, and returns to step S1. The application can solve the problems of NLOS error, single model unable to match the environment, and inaccurate noise statistical characteristics in the indoor environment affecting the positioning accuracy, and improve the positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of indoor positioning technology, and more particularly to an interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering. Background Technology

[0002] In the era of intelligent technology, location services are closely related to daily life, such as real-time positioning and trajectory tracking of rescue personnel in the rescue field, equipment tracking and special patient management in the medical field, and over / understaffing alarms and intelligent supervision of inspection work in the industrial field. Traditional satellite positioning technology can achieve good positioning results in outdoor environments. However, in indoor environments, its performance is significantly reduced due to signal obstruction by various obstacles.

[0003] Indoor environments are often a mixture of line-of-sight (LOS) and non-line-of-sight (NLOS) conditions. NLOS errors caused by signal transmission under NLOS conditions are the main cause of inaccurate indoor positioning. Existing technologies address this problem primarily through two approaches: identifying NLOS errors and suppressing NLOS errors. Methods based on identifying NLOS errors first use an identification algorithm to remove data containing NLOS errors, then use the remaining LOS data for position calculation. However, this method experiences significant performance degradation in scenarios with dense NLOS conditions. Methods based on suppressing NLOS errors directly process the system input data, suppressing NLOS errors through various means during position calculation. However, these algorithms mostly use a single observation model to match the environment, failing to consider the decrease in positioning accuracy caused by frequent switching between LOS and NLOS conditions when the target moves. Furthermore, the inaccurate statistical characteristics of noise due to the variable indoor environment also contribute to decreased positioning accuracy. Summary of the Invention

[0004] This invention provides an indoor positioning method based on an Adaptive Unscented Kalman Filter (IMM-AUKF). It addresses the problems of NLOS error, the inability of a single model to match the environment, and the inaccurate noise statistics affecting positioning accuracy in indoor environments.

[0005] The technical means employed in this invention are as follows:

[0006] An interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering is implemented using a positioning calculation terminal device. The positioning calculation terminal device includes a main control module, a ranging sensor module, a display module, and a power supply module. The power supply module provides power to the positioning calculation terminal device. The display module displays the position and trajectory information of the target to be located. The ranging sensor module includes beacon nodes and routing nodes, used to measure the distance between the beacon nodes and the target to be located and provide this information to the main control module. The main control module includes a coordinator, a server, and a host computer, used to calculate the position and trajectory of the target to be located based on the distance between the beacon nodes and the target to be located provided by the ranging sensor module.

[0007] The method includes the following steps:

[0008] S1. The ranging sensor module obtains the flight time through communication between the beacon node and the target to be located based on the flight time, calculates the distance information between the beacon node and the target to be located based on the flight time, and provides it to the main control module through the routing node.

[0009] S2. The main control module sends the distance information from the ranging sensor module to the server for storage and processing through the coordinator, and finally calculates the position coordinates in the host computer using the IMM-AUKF algorithm.

[0010] S3. The display module stores and displays the position coordinates and movement trajectory of the target to be located calculated in the main control module, and then proceeds to step S1.

[0011] Compared with the prior art, the present invention has the following advantages:

[0012] In this invention, firstly, based on the different characteristics of ranging errors in indoor LOS and NLOS environments, two parallel UKFs are set up in the IMM to simultaneously filter the ranging values. Then, using the corresponding model probabilities as weighting factors, the results of the LOS and NLOS filtering models are weighted and accumulated to obtain the position coordinates, ultimately suppressing the NLOS error. Furthermore, an adaptive factor based on the innovation vector is introduced into the UKF to reduce the impact of inaccurate noise statistical characteristics on positioning accuracy by adjusting the covariance matrix of the process noise in real time. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating an IMM algorithm based on adaptive UKF provided by the present invention.

[0015] Figure 2 This is a flowchart illustrating an adaptive UKF algorithm provided by the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] like Figure 1 As shown, this invention provides an interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering, which is implemented by a positioning calculation terminal device. The positioning calculation terminal device consists of a main control module, a ranging sensor module, a display module, and a power supply module. The power supply module provides power support for the positioning calculation terminal device. The display module displays the position and trajectory information of the target to be located. The ranging sensor module consists of beacon nodes and routing nodes, measuring the distance between the beacon nodes and the target to be located and providing this distance to the main control module. The main control module is the core unit of the positioning calculation terminal device, consisting of a coordinator, a server, and a host computer, and calculates the position and trajectory of the target to be located based on the ranging information provided by the ranging sensor module.

[0018] The processing steps of the IMM-AUKF-based indoor positioning method are as follows:

[0019] S1. The ranging sensor module obtains the flight time through communication between the beacon node and the target to be located. Based on the flight time, the distance information between the beacon node and the target to be located is calculated and provided to the main control module through the routing node.

[0020] S2. The main control module sends the ranging values ​​from the ranging sensor module to the server for storage and processing via a coordinator. Finally, the position coordinates are calculated in the host computer using the IMM-AUKF algorithm. The specific calculation process of the IMM-AUKF algorithm is as follows:

[0021] S201. Based on the different characteristics of ranging errors under LOS and NLOS environments, UKF filters based on LOS and NLOS are established respectively, and they are combined with the IMM algorithm as sub-filters.

[0022] In this embodiment of the invention, ranging models are constructed based on LOS and NLOS respectively in an indoor LOS / NLOS mixed environment.

[0023] The LOS-based ranging model is as follows:

[0024]

[0025] in, This represents the actual distance between the coordinates of the target to be located and the coordinates of the m-th beacon node. This refers to the noise generated by the measurement system under LOS.

[0026] The NLOS-based ranging model is as follows:

[0027]

[0028] In the formula, Noise generated by the measurement system under NLOS.

[0029] The system equations for the target to be located are:

[0030]

[0031] Let be the state vector of the target to be located. This represents process noise. The state transition matrix F, control matrix G, and measurement noise ω are also considered. m (k) is shown below:

[0032]

[0033]

[0034]

[0035] In the matrix above, T represents the sampling period.

[0036] This invention provides an interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering, which mainly consists of two parts: adaptive UKF and IMM. Adaptive UKF adjusts the covariance matrix of process noise in real time using an adaptive factor based on the innovation vector, thereby reducing the impact of inaccurate noise statistics on positioning accuracy. IMM constructs two UKFs to match the mixed indoor environment and uses model probabilities as weighting factors to weight and accumulate the filtering results, ultimately suppressing NLOS errors.

[0037] S202. Calculate the mixed probability based on the model probability and model transition probability at the initial time.

[0038] The specific method for calculating the mixture probability is as follows:

[0039]

[0040]

[0041] Where i, j = 1, 2. π ij This represents the model transition probability in a Markov chain. i (k-1) represents the model probability of model i at time k-1. The mixture probability of model i and model j at time k-1 is denoted by u. i|j (k-1|k-1) represents this. It is a normalization constant.

[0042] S203. Perform interactive calculations based on the state estimate and the mixture probability to obtain the initial filtering values ​​for the LOS and NLOS filters.

[0043] The specific method for performing interactive calculations is as follows:

[0044]

[0045]

[0046] in, P i (k-1) represents the state estimate and corresponding covariance matrix of model i at time k-1, respectively. Then... P 0j (k-1) represents the state estimate and the corresponding covariance matrix after the mixed calculation at time k-1.

[0047] S204. The distance information between the beacon node and the target to be located is calculated by the ranging sensor module using a TOF-based method.

[0048] S205. Using the distance information and the initial filtering value as input to the filter, perform both LOS-based adaptive UKF filtering and NLOS-based adaptive UKF filtering simultaneously. Adaptive UKF filtering can be divided into two parts: UKF filtering and adaptive factor update. The specific process is as follows:

[0049] S2051. Generate the Sigma point set and its weights using the UT transform based on the initial filtering value. Specifically, the process of generating the Sigma point set using the UT transform is shown in formula (18). The process of generating the weights is shown in formulas (23) and (24).

[0050] S2052. Calculate the one-step prediction value based on the Sigma point set. See formula (19) for the specific prediction process.

[0051] S2053. Calculate the one-step predicted value of the system state variables based on the initial value of the filter. See formula (20) and formula (22) for the specific process.

[0052] S2054. Substitute the Sigma point set into the observation equation to obtain the predicted values ​​of the observations.

[0053] The observation equation is:

[0054]

[0055] The process of obtaining the predicted values ​​of the observed quantities is shown in formula (25).

[0056] S2055. The mean of the new observations is obtained by weighting. See formula (26) for the specific process.

[0057] S2056. Obtain the innovation vector by subtracting the distance information from the system prediction value. See formula (30) for the specific process.

[0058] S2057. Calculate the covariance matrix of the innovation vector. See formula (34) for the specific process.

[0059] S2058. Construct an adaptive factor based on the covariance matrix of the innovation vector and the UKF posterior second-order statistical properties, and introduce it into step S2053 for calculation. The construction method of the adaptive factor is shown in formulas (33) to (35). The exact form of the constructed adaptive factor is shown in formula (36).

[0060] S2059. Calculate the Kalman filter gain based on the variance of the new observations. The calculation formula is shown in formula (29).

[0061] S20510. Calculate and output the final estimates of the state vector and covariance based on the Kalman filter gain. The state vector is given by formula (31), and the covariance is given by formula (32).

[0062] S206. Calculate the likelihood function of each model based on the covariance and innovation vector of the adaptive UKF output. The formula for calculating the likelihood function is given in formula (11).

[0063] S207. Update the model probability of each model based on the likelihood function and standardization constant of each model.

[0064] The specific method for updating the model probabilities is as follows:

[0065]

[0066]

[0067]

[0068] Among them, vj (k) represents the innovation vector obtained by UKF filtering of model j at time k. The specific calculation process is as follows: P represents the distance predicted by the system at time k, as shown in equation (25). yy The covariance matrix obtained after UKF filtering is shown in equation (26), det(P yy ) represents finding the covariance matrix P yy The value of the determinant of Λ. j (k) represents the likelihood function value of model j at time k.

[0069] S208. Based on the statistical characteristics of the above model probability and the adaptive UKF output, perform a fusion calculation to obtain the final estimate.

[0070] The specific fusion estimation method is as follows:

[0071]

[0072]

[0073] in, P(k) represents the final state estimate of the target to be located at time k and the corresponding covariance matrix, respectively.

[0074] S209. The final position coordinates are calculated by multiplying the final estimate above with the observation matrix.

[0075] The specific location coordinates are obtained as follows:

[0076]

[0077]

[0078] Where B represents the observation matrix and L(k) represents the two-dimensional coordinates of the target to be located at time k.

[0079] As a preferred embodiment of the present invention, this embodiment combines the above-mentioned adaptive UKF operation with... Figure 2 The specific operation steps can be represented as follows:

[0080] 1. Prediction Section

[0081]

[0082] in, P(k-1) represents the pre-defined state vector and covariance at time k-1. n is the system dimension. λ = α 2(n+κ)-n is a parameter used to obtain the Sigma point set, mainly controlled by the value of α to determine the degree of dispersion of the Sigma points. ξ (i) (k-1) represents the i-th sampling point at time k-1.

[0083] ξ i (k)=Fξ i (k-1), i=0,1,…2n (19)

[0084]

[0085] in, W is the one-step prediction of the state vector. i m The weighting factor is the mean; the specific calculation process is shown in equation (23).

[0086] E(k-1)=ψ(k-1)GQ(k-1)G T ψ T (k-1) (21)

[0087]

[0088] Where P(k|k-1) is the one-step prediction of the covariance, and W i c The variance is a weighting factor, and the specific calculation process is shown in equation (24). E(k-1) is an intermediate variable introduced into the simplified formula and has no special meaning.

[0089] Q(k-1) is the covariance matrix of the process noise at time k-1. ψ(k-1) is the adaptive matrix constructed by the adaptive factor at time k-1.

[0090]

[0091]

[0092] β is the parameter for constructing the variance weighting factor, and its optimal value is 2 when the noise follows a Gaussian distribution.

[0093] χ i (k)=h(ξ i (k)) (25)

[0094] χ i (k) represents the predicted value of the observed quantity. h(·) represents the observation equation.

[0095]

[0096] This represents the mean of the observed values.

[0097]

[0098]

[0099] P yy (k), P xy (k) represent the system's autocovariance and crosscovariance, respectively. R(k) is the covariance of the measurement noise.

[0100] 2. Updated section

[0101] This section is primarily used for updating the Kalman filter gain, innovation vector, state vector, and covariance. Specifically:

[0102]

[0103]

[0104]

[0105] P(k)=P(k|k-1)-K(k)P yy (k)K(k) T (32)

[0106] Where K(k) is the Kalman filter gain at time k, and ν(k) is the innovation vector at time k.

[0107] Z m (k) represents the distance between the m-th beacon node and the target to be located at time k. P(k) represents the state vector and covariance at time k.

[0108] 3. Calculate the adaptive factor

[0109] Δ(k)=E{v(k)v(k) T}-P yy (k) (33)

[0110]

[0111]

[0112]

[0113] Δ(k) represents the covariance of the error caused by inaccurate statistical characteristics of noise. V(k) represents the covariance of the innovation vector, where ρ is the forgetting factor, which generally satisfies 0 < ρ ≤ 1. Let denot be the adaptive factor at time k, and η be the pre-set decision threshold. λ1 and λ2 are forgetting factors, whose values ​​can be adjusted to control the steady-state estimation performance and the accuracy of the positioning results. ψ(k) is the adaptive matrix composed of the adaptive factors.

[0114] S3. The display module stores and displays the position coordinates and movement trajectory of the target to be located calculated in the main control module, and then returns to S1 to repeat the above process.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering, implemented using a positioning calculation terminal device, wherein the positioning calculation terminal device includes a main control module, a ranging sensor module, a display module, and a power supply module; the power supply module provides power support for the positioning calculation terminal device; the display module displays the position and trajectory information of the target to be located; the ranging sensor module includes beacon nodes and routing nodes, used to measure the distance between the beacon nodes and the target to be located and provide it to the main control module; the main control module includes a coordinator, a server, and a host computer, used to calculate the position and trajectory of the target to be located based on the distance between the beacon nodes and the target to be located provided by the ranging sensor module; Its features are, The method includes the following steps: S1. The ranging sensor module obtains the flight time through communication between the beacon node and the target to be located based on the flight time, calculates the distance information between the beacon node and the target to be located based on the flight time, and provides it to the main control module through the routing node. S2. The main control module sends the distance information from the ranging sensor module to the server for storage and processing via a coordinator. Finally, the position coordinates are calculated in the host computer using the IMM-AUKF algorithm, including: S201. Establish a UKF filter based on LOS based on the ranging error under LOS environment, and establish a UKF filter based on NLOS based on the ranging error under NLOS environment. Combine the two filters with the IMM algorithm as sub-filters. S202. Calculate the mixture probability based on the model probability and model transition probability at the initial time. S203. Perform interactive calculations based on the state estimate and the mixture probability to obtain the initial filter values ​​for the LOS and NLOS filters; S204. The distance information between the beacon node and the target to be located is calculated by the ranging sensor module using a TOF-based method. S205. The distance information and the initial filtering value are used as the input of the filter, and adaptive UKF filtering based on LOS and adaptive UKF filtering based on NLOS are performed simultaneously. S206. Calculate the likelihood function of each model based on the covariance and innovation vector of the adaptive UKF output; S207. Update the model probability of each model based on the likelihood function and standardization constant of each model; S208. Perform a fusion calculation based on the statistical characteristics of the model probability and the adaptive UKF output to obtain the final estimate; S209. The final position coordinates are calculated by multiplying the final estimate with the observation matrix. S3. The display module stores and displays the position coordinates and movement trajectory of the target to be located calculated in the main control module, and then proceeds to step S1.

2. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The adaptive UKF filtering specifically includes: S2051. Generate the Sigma point set and its weights using UT transformation based on the initial filtering values; S2052. Calculate the one-step prediction value based on the Sigma point set; S2053. Calculate the one-step predicted value of the system state variables based on the initial value of the filter; S2054. Substitute the Sigma point set into the observation equation to obtain the predicted values ​​of the observations; S2055. The mean of the new observations is obtained by weighting. S2056. Obtain the innovation vector by subtracting the distance information from the system prediction value; S2057. Calculate the covariance matrix of the innovation vector; S2058. Construct an adaptive factor based on the covariance matrix of the innovation vector and the UKF posterior second-order statistical properties, and introduce it into step S2053 for calculation; S2059. Calculate the Kalman filter gain based on the variance of the new observations; S20510. Calculate and output the final estimates of the state vector and covariance based on the Kalman filter gain.

3. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The distance between the beacon node and the target to be located is calculated based on a LOS-based ranging model, which is as follows: (1) in, Indicates the coordinates of the target to be located and the first... m The actual distance between the coordinates of each beacon node. Noise generated by the measurement system under LOS; The NLOS-based ranging model is as follows: (2) In the formula, Noise generated by the measurement system under NLOS; The system equations for the target to be located are: (3) Let be the state vector of the target to be located. ( k ( ) represents process noise, state transition matrix F, control matrix G, and measurement noise. As shown below: (4) (5) (6) in, T The sampling period.

4. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The mixture probability is calculated based on the initial model probability and the model transition probability, including: (7) (8) in, i , j =1,2, This represents the model transition probability in a Markov chain. express k -1 time model i The model probability, For the model i and model j exist k The mixture probability at time -1 It is a normalization constant.

5. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The initial filter values ​​for the LOS and NLOS filters are obtained through interactive computation based on the state estimate and mixture probability, including: (9) (10) in, , The models are respectively i exist k The state estimate and corresponding covariance matrix at time -1 , They are respectively in k The state estimate and corresponding covariance matrix after mixed calculation at time -1.

6. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The model probabilities of each model are updated based on their likelihood functions and standardization constants, including: (11) (12) (13) in, Representation Model j exist k The innovation vector obtained after UKF filtering at each time step is calculated as follows: , for k The distance value predicted by the time system, This is the covariance matrix obtained after UKF filtering. This represents finding the covariance matrix. The value of the determinant, Representation Model j exist k The likelihood function value at time t.

7. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The final estimate is obtained by fusing the model probability with the statistical characteristics of the adaptive UKF output, including: (14) (15) in, , The target to be located is in k The final state estimate and corresponding covariance matrix at time t.

8. The interactive multi-model indoor positioning method based on adaptive unscented Kalman filtering according to claim 1, characterized in that, The final position coordinates are calculated by multiplying the final estimate with the observation matrix, including: (16) (17) in, Represents the observation matrix. Indicates that the target to be located is in k Two-dimensional coordinates at time.

Images