Indoor positioning method and device under occlusion environment

Abstract

The application discloses an indoor positioning method and device under shielding environment, and relates to the technical field of indoor positioning, and comprises the following steps: obtaining a distance calculation value and a relative motion speed calculation value of a mobile target and a base station according to sound signal information of the mobile target position at a certain moment; constructing a fitting error matrix by using a difference between a distance observation value of the mobile target and the base station at the moment and the distance calculation value and a difference between a relative motion speed observation value and the relative motion speed calculation value; using a line-of-sight information matrix obtained by removing non-line-of-sight measurement information by using a line-of-sight identification algorithm based on Gaussian Bayesian dynamic training as a parameter matrix of a weighted least square problem based on an identification-removal strategy; obtaining a target problem expression; solving the target problem expression by using a Levenberg-Marquardt algorithm; and obtaining a position estimation result of a position sequence frame at a position estimation moment. The method can realize high-precision position estimation in a shielding environment or a low base station deployment density environment.

Description

Technical Field

[0001] This invention relates to the field of indoor positioning technology, and more specifically to an indoor positioning method and apparatus in an obstructed environment. Background Technology

[0002] High-precision location information is crucial for smart mobile terminals to achieve various applications and is the cornerstone for the future development and promotion of intelligent systems. Over the past decade, scholars have proposed many indoor positioning systems based on technologies such as Bluetooth, WiFi, ultra-wideband, inertial navigation, geomagnetism, and ultrasound. Compared with other technologies, sound-based indoor positioning methods use consumer-grade speakers and microphones as measurement elements, which have high measurement accuracy and anti-electromagnetic interference capabilities, and are compatible with the vast majority of smart mobile terminals. This makes it the most competitive solution for smart mobile terminals to achieve high-precision indoor positioning and navigation in complex indoor environments.

[0003] The main technologies of indoor sound positioning systems include positioning methods based on Time of Arrival (TOA), Time Difference of Arrival (TDOA), Direction of Arrival (DOA), and fingerprint matching. TOA-based positioning systems require three pre-set base stations (anchor nodes) and high-precision clock synchronization between the base stations and the agent, typically offering higher positioning accuracy and stability. However, in complex indoor environments with occlusion, the line-of-sight path between the measured location and the base station often disappears due to walls or objects, introducing a significant positive bias in time delay estimation. This severely limits and weakens the accuracy and stability of indoor sound positioning systems. The results of the Microsoft Indoor Positioning Competition indicate that positioning under occlusion is a major obstacle to the application and promotion of distance-based indoor positioning technologies. Achieving high-precision sound positioning for mobile terminals in complex indoor environments with occlusion remains an open problem.

[0004] In the field of sound-based indoor positioning, Lin Chi-mak and Tomonari Furukawa proposed a low-frequency sound occlusion positioning method based on indoor maps or indoor environmental information, taking into account the characteristics of indoor sound propagation. This method estimates the target location by calculating the mirror position of the occluded base station, solving the occlusion positioning problem in ideal environments. In 2018, Niranjini Rajagopal et al. from Carnegie Mellon University achieved target positioning under strong occlusion conditions based on indoor planar map information and assumptions about line-of-sight and occlusion information within the entire planar area, improving positioning performance from 4 in real-world scenarios. The value of 8 (m) was increased to 1 m.

[0005] In practical scenarios, position estimation methods based on least squares are the most widely used. To address ranging bias caused by occlusion, weighted least squares can be used to suppress this bias. Pi-Chun Chen proposed a residual-weighted least squares method, Rwgh (Residual Weighting Algorithm). First, the residuals are calculated under different measurement combinations, and the weight vector is chosen as a function of the inverse of the residuals to estimate the position. However, this method has limitations on the ratio of line-of-sight to non-line-of-sight measurements and requires a minimum of 3 line-of-sight measurements; otherwise, ideal results cannot be obtained.

[0006] Based on the identification-rejection strategy, the statistical characteristics of historical measurement data are used to identify abnormal measurements. Only measurements with high reliability are selected for location estimation, which is a relatively effective occlusion localization strategy. However, existing occlusion localization methods based on the occlusion-rejection strategy require a high base station (beacon) deployment density. At the same time, there are limitations on the ratio of line-of-sight to non-line-of-sight measurements, and a minimum of 3 line-of-sight measurements are required. Otherwise, ideal results cannot be obtained. Existing distance-based localization technology treats the location estimation at different times as an independent process. The target's motion information is not fully utilized, the localization stability is not high, and the localization accuracy in real-world scenarios is poor. Summary of the Invention

[0007] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, a first aspect of this invention proposes an indoor positioning method in an obstructed environment, comprising:

[0008] Based on the acoustic signal information of the moving target's location at a certain moment, the calculated distance and relative speed between the moving target and the base station are obtained.

[0009] The fitting error matrix, constructed using the differences between the observed and calculated distance values ​​between the moving target and the base station at that moment, and the observed and calculated relative motion speed values, together with the line-of-sight information matrix obtained by eliminating non-line-of-sight measurement information using a Gaussian Bayes dynamic training line-of-sight recognition algorithm, is used as the parameter matrix for the weighted least squares problem based on the recognition-elimination strategy, thus obtaining the expression for the target problem.

[0010] The Levenberg-Marquardt algorithm is used to solve the objective problem expression and obtain the position estimation results of the position sequence frames at the time of the position to be estimated. The initial values ​​of the Levenberg-Marquardt algorithm include the initial position sequence frames to be estimated that only represent distance information, the short-time position sequence frames to be estimated that simultaneously represent distance information and relative velocity, and the short-time position sequence frame length set. After determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the iteration of the Levenberg-Marquardt algorithm is terminated by the fitting residual and the preset maximum frame length. During the iteration process, the length of the sequence frames is dynamically adjusted to obtain the optimal sequence frame length.

[0011] Furthermore, based on the acoustic signal information of the moving target's location at a certain moment, the calculated distance and relative velocity between the moving target and the base station are obtained, including:

[0012] Obtain a composite signal comprising two hyperbolic frequency modulated signals;

[0013] The first and second delay information of the composite signal are obtained by two matched filters, respectively.

[0014] Based on the first and second time delay information, the estimated time delay and the estimated Doppler factor of the composite signal are obtained.

[0015] The distance between the moving target and the base station and the relative speed are calculated by multiplying the estimated time delay and the estimated Doppler factor with the speed of sound, respectively.

[0016] Furthermore, the fitting error matrix constructed using the differences between the observed and calculated distance values ​​between the moving target and the base station at that moment, and between the observed and calculated relative motion speed values, along with the line-of-sight information matrix obtained by eliminating non-line-of-sight measurement information using a Gaussian Bayes-based dynamic training line-of-sight recognition algorithm, serves as the parameter matrix for the weighted least squares problem based on the recognition-elimination strategy, including:

[0017] Using the relative motion velocity observation value and distance observation value between the moving target and the current line-of-sight base station at any given time, the coordinates of the current base station, and the position to be estimated of the moving target at the current time as parameters, the distance calculation value is obtained by using the first preset formula, and the relative velocity observation value is obtained by using the second preset formula.

[0018] Based on the observed and calculated distance values ​​between the moving target and the base station, as well as the observed and calculated relative velocity values, the distance error values ​​of the distance error elements and the relative velocity error values ​​of the velocity error elements of the fitted error matrix are obtained.

[0019] Based on the line-of-sight information matrix, fitting error matrix, and pre-defined weighting matrix of the weighted least squares problem based on the recognition-removal strategy obtained by eliminating non-line-of-sight measurement information using the line-of-sight recognition algorithm based on Gaussian Bayes dynamic training, this algorithm is used.

[0020] Furthermore, based on the observed and calculated distance values ​​between the moving target and the base station, and the observed and calculated relative velocity values, the distance error values ​​of the distance error elements and the relative velocity error values ​​of the velocity error elements of the fitted error matrix are obtained, including:

[0021] The distance is calculated according to the first preset formula, and the relative velocity is calculated according to the second preset formula. The expressions of the first preset formula and the second preset formula are as shown in expressions (1) and (2):

[0022]

[0023] in, Indicates the moving target is in The position to be estimated at a given time. Indicates the moving target is in The position to be estimated at a given time. This indicates the coordinates of the current line-of-sight base station. This indicates the period for synchronization of multiple base stations. This represents the calculated distance value. The calculated value representing relative velocity;

[0024] The distance error value is obtained according to expression (3) and used as an element of the distance error in the fitting error matrix;

[0025]

[0026] in, Represents distance observation value, Indicates distance error Element;

[0027] The relative velocity error value is obtained according to expression (4) and used as an element of the relative velocity error in the fitting error matrix;

[0028]

[0029] in, Represents the observed relative velocity values. Indicates speed error Element;

[0030] Based on distance error and speed error , obtained the The time estimation error;

[0031] according to T The estimation error of +1 short-time position sequence is used to obtain the fitting error matrix. Its expression is shown in formula (5);

[0032]

[0033] in T This indicates the length of the short-time position sequence.

[0034] Furthermore, the expression for the pre-defined weighting matrix of the weighted least squares problem based on the identification-elimination strategy includes:

[0035]

[0036] in, , , , , , For the fitting error vector, For weighted matrices, This is the line-of-sight information matrix. Represents a diagonal matrix. It is a distance-weighted matrix. The relative velocity weighting matrix is ​​obtained by changing... and Adjusting distance and relative velocity observations can alter the positioning results. It is a line-of-sight information matrix, and has .

[0037] Furthermore, after determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the Levenberg-Marquardt algorithm is terminated using the fitting residual and the preset maximum short-time position sequence frame length as the iteration conditions. During the iteration process, the sequence frame length is dynamically adjusted to obtain the optimal sequence frame length, including:

[0038] Based on the line-of-sight information matrix, it is determined that there are at least two base stations closest to the moving target;

[0039] The short-time position sequence frame length values ​​in the acquired sequence frame length set, arranged in ascending order;

[0040] Based on the comparison results between the short-time position sequence frame length value obtained in each iteration and the current position time value to be estimated, it is determined whether to reduce the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation.

[0041] Based on the first comparison result of the fitting residual values ​​of two adjacent fitting error matrices and the second comparison result of the actual short-time position sequence frame length value and the maximum short-time position sequence frame length value, determine whether to output the position estimation result of the position sequence frame at the time to be estimated.

[0042] Furthermore, based on the comparison results of the short-time position sequence frame length value obtained in each iteration with the current position time value to be estimated, it is determined whether to reduce the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation, including:

[0043] In each iteration, if the obtained short-time position sequence frame length value is greater than or equal to the estimated position time value, then the short-time position sequence frame length value calculated in each iteration is reduced by 1 and assigned to the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation.

[0044] If the short-time position sequence frame length value obtained in each iteration is less than the time value of the position to be estimated, then the obtained short-time position sequence frame length value is used as the actual short-time position sequence frame length value used in the iterative calculation of the Levenberg-Marquardt algorithm.

[0045] Furthermore, based on the line-of-sight information matrix, it is determined that there are at least two base stations closest to the moving target, including:

[0046] Obtain the current line-of-sight information matrix;

[0047] Then, determine whether the norm of the line-of-sight information matrix is ​​greater than 2. If it is less than 2, select the base station closest to the moving target from the obscured base stations, set the element corresponding to the base station in the line-of-sight information matrix to 1, and continue to determine whether the norm of the line-of-sight information matrix is ​​greater than 2. After it is greater than 2, it is determined that there are at least two base stations closest to the moving target.

[0048] Another aspect of the present invention provides an indoor positioning device in an obstructed environment, comprising:

[0049] The acquisition module is used to obtain the calculated distance and relative speed between the moving target and the base station based on the acoustic signal information of the moving target's location at a certain moment.

[0050] The processing module is used to construct a fitting error matrix by using the difference between the observed and calculated distance values ​​between the moving target and the base station at that moment, and the difference between the observed and calculated relative motion speed values, and the line-of-sight information matrix obtained by using a line-of-sight recognition algorithm based on Gaussian Bayes dynamic training to eliminate non-line-of-sight measurement information, as the parameter matrix of the weighted least squares problem based on the recognition-elimination strategy, to obtain the expression of the target problem.

[0051] The output module is used to solve the objective problem expression using the Levenberg-Marquardt algorithm to obtain the position estimation results of the position sequence frames at the time of the position to be estimated. The initial values ​​of the Levenberg-Marquardt algorithm include the initial position sequence frames to be estimated that only represent distance information, the short-time position sequence frames to be estimated that simultaneously represent distance information and relative velocity, and the short-time position sequence frame length set. After determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the iteration of the Levenberg-Marquardt algorithm is terminated by the fitting residual and the preset maximum frame length. During the iteration process, the length of the sequence frames is dynamically adjusted to obtain the optimal length of the sequence frames.

[0052] This invention provides an indoor positioning method and apparatus for obstructed environments, which, compared with the prior art, have the following advantages:

[0053] This invention addresses the need for high-precision indoor positioning of smart mobile terminals in indoor occlusion scenarios. By simultaneously estimating the distance and relative velocity of sound signals, and using line-of-sight measurements based on distance and relative velocity, multiple positions in a short-time position sequence are jointly estimated. The cumulative error introduced by velocity integration is used to improve the overall position estimation accuracy, thereby achieving high-precision indoor sound occlusion positioning for smart mobile terminals in indoor occlusion scenarios. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0055] Figure 1 A flowchart illustrating an indoor positioning method in an obstructed environment, as provided in an embodiment of the present invention;

[0056] Figure 2 A flowchart illustrating the positioning process of an indoor positioning method in an obstructed environment, provided by an embodiment of the present invention.

[0057] Figure 3 An algorithm architecture diagram of an indoor positioning method in an obstructed environment provided by an embodiment of the present invention;

[0058] Figure 4 A flowchart illustrating an indoor positioning method in an obstructed environment, provided as an embodiment of the present invention.

[0059] Figure 5 A numerical simulation comparison of position estimation for an indoor positioning method under obstructed conditions, provided in an embodiment of the present invention;

[0060] Figure 6 A numerical simulation performance comparison chart of an indoor positioning method under obstructed conditions provided by an embodiment of the present invention;

[0061] Figure 7 Comparison chart of numerical simulation performance of an indoor positioning method in an obstructed environment provided by an embodiment of the present invention;

[0062] Figure 8 An indoor positioning method in an obstructed environment is provided as an embodiment of the present invention. , Performance comparison chart of simulation results under different numbers of base stations;

[0063] Figure 9 An experimental scenario diagram of an indoor positioning method under obstructed conditions provided in an embodiment of the present invention;

[0064] Figure 10 A comparative diagram of experimental results for an indoor positioning method under obstructed conditions provided in an embodiment of the present invention;

[0065] Figure 11 A comparative experimental performance chart of an indoor positioning method under obstructed conditions provided in an embodiment of the present invention;

[0066] Figure 12 This is a structural diagram of an indoor positioning device in an obstructed environment, provided as an embodiment of the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0068] This specification provides the operational steps for the methods described in the embodiments or flowcharts, but may include more or fewer operational steps based on conventional or non-inventive labor. In actual system or server product execution, the methods shown in the embodiments or drawings may be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0069] Figure 1 A flowchart of the method provided by the present invention, such as Figure 1 As shown, the method includes:

[0070] S101. Based on the acoustic signal information of the moving target's location at a certain moment, obtain the calculated distance and relative speed between the moving target and the base station.

[0071] It should be noted that the acoustic signal is a composite signal constructed based on two hyperbolic frequency modulated (HFM) signal components. The calculation process for calculating the distance between the moving target and the base station using the received acoustic signal is based on the patent with reference to patent number CN202110566504.X, which will not be repeated here.

[0072] S102. The fitting error matrix constructed using the difference between the observed and calculated distance values ​​between the moving target and the base station at that moment, and the difference between the observed and calculated relative motion speed values, together with the line-of-sight information matrix obtained by eliminating non-line-of-sight measurement information using the line-of-sight recognition algorithm based on Gaussian Bayes dynamic training, is used as the parameter matrix of the weighted least squares problem based on the recognition-elimination strategy to obtain the expression of the target problem.

[0073] In this invention, the moment refers to a certain moment, which is one of multiple moment points. The distance calculation value and the relative motion speed calculation value between the moving target and the base station are obtained through S102. The observed value is the actual measured value, including the estimated distance and relative motion speed between the moving target and the base station. The Gaussian Bayes dynamic training line-of-sight recognition algorithm is an algorithm for obtaining the line-of-sight matrix, which can suppress non-line-of-sight information.

[0074] Steps S101 and S102 are the process of obtaining basic information. In these two steps, the parameter matrix of the weighted least squares problem based on the identification-elimination strategy is established, thus laying the foundation for the mathematical objective function of the weighted least squares problem based on the identification-elimination strategy.

[0075] S103. Using the Levenberg-Marquardt algorithm, solve the objective problem expression to obtain the position estimation results of the position sequence frames at the time of the position to be estimated. The initial values ​​of the Levenberg-Marquardt algorithm include the initial position sequence frames to be estimated that only represent distance information, the short-time position sequence frames to be estimated that simultaneously represent distance information and relative velocity, and the short-time position sequence frame length set. After determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the fitting residual and the preset maximum frame length are used as the termination conditions for the Levenberg-Marquardt algorithm iteration. During the iteration process, the length of the sequence frames is dynamically adjusted to obtain the optimal length of the sequence frames.

[0076] It should be noted that the Levenberg-Marquardt algorithm is a commonly used method for solving weighted least squares problems, and the setting of its initial values ​​is crucial.

[0077] In this invention, the initial values ​​include an initial position sequence frame to be estimated that only represents distance information, a short-time position sequence frame to be estimated that simultaneously represents distance information and relative velocity, and a set of short-time position sequence frame lengths.

[0078] In this invention, a preset fitting residual is also included, which is to prevent the loop from being exited during the first iteration.

[0079] In the embodiments provided by the present invention, the line-of-sight recognition method based on Gaussian Bayes dynamic training is used to identify the line-of-sight measurement information between the target and the base station, ensuring that the number of line-of-sight base stations at the current time is not less than 2, and the fitting residual and the preset maximum frame length are used as the termination conditions for the Levenberg-Marquardt algorithm iteration. During the iteration process, the length of the sequence frame is dynamically adjusted to obtain the optimal length of the sequence frame.

[0080] It should be noted that the Levenberg-Marquardt algorithm is used for solving this problem, and its core lies in calculating the Jacobian matrix and the Hessell matrix. This varies with frame length. With the increase in the number of elements, the dimensions of the two types of matrices increase rapidly, and the memory required for computation and the number of floating-point calculations also increase dramatically, which in turn increases the power consumption and cost of the overall system, but the accuracy of position estimation will also be improved.

[0081] In embodiments of the present invention, the positioning principle framework is as follows: Figure 2 As shown, in the first Short-time location sequence frames at time t, by simultaneous estimation The position at a given moment, i.e. To suppress distance and relative velocity measurement errors caused by occlusion, the frame position is adjusted. Estimating it as an unknown quantity can reduce the cumulative error of velocity and improve the overall estimation accuracy.

[0082] In summary, as Figure 3 As shown, in the algorithm architecture of this embodiment of the invention Figure 3 In this study, line-of-sight recognition, distance and relative velocity estimation are performed on the received acoustic signals. An information matrix is ​​constructed based on a Gaussian Bayes dynamic training method for line-of-sight recognition. Based on the Levenberg-Marquardt (LM) algorithm for fast solution of weighted nonlinear least squares, and using the fitting residual and the preset maximum frame length as the termination condition of the Levenberg-Marquardt algorithm iteration, the length of the sequence frame is dynamically adjusted during the iteration process to obtain the optimal length of the sequence frame. While ensuring low computational cost, the overall estimation accuracy of each position in the sequence frame is improved.

[0083] In one possible implementation, based on the acoustic signal information of the moving target's location at a certain moment, the calculated distance and relative speed between the moving target and the base station are obtained, including:

[0084] Obtain a composite signal comprising two hyperbolic frequency modulated signals;

[0085] The first and second delay information of the composite signal are obtained by two matched filters, respectively.

[0086] Based on the first and second time delay information, the estimated time delay and the estimated Doppler factor of the composite signal are obtained.

[0087] The distance between the moving target and the base station and the relative speed are calculated by multiplying the estimated time delay and the estimated Doppler factor with the speed of sound, respectively.

[0088] In one embodiment of the present invention, the distance between the moving target and the base station and the relative speed are calculated, as described in patent CN202110566504.X.

[0089] This invention makes a first-order assumption about the kinematic model of the target, which is too large. The value of the sequence frame length may actually reduce the accuracy of the position estimation, so it is necessary to choose an appropriate sequence frame length. .

[0090] In one possible implementation, a fitting error matrix constructed using the differences between the observed and calculated distance values ​​between the moving target and the base station at that moment, and between the observed and calculated relative motion speed values, along with a line-of-sight information matrix obtained by eliminating non-line-of-sight measurement information using a Gaussian Bayes-based dynamic training line-of-sight recognition algorithm, serves as the parameter matrix for a weighted least squares problem based on a recognition-elimination strategy, including:

[0091] Using the relative motion velocity observation value and distance observation value between the moving target and the current line-of-sight base station at any given time, the coordinates of the current base station, and the position to be estimated of the moving target at the current time as parameters, the distance calculation value is obtained by using the first preset formula, and the relative velocity observation value is obtained by using the second preset formula.

[0092] Based on the observed and calculated distance values ​​between the moving target and the base station, as well as the observed and calculated relative velocity values, the distance error values ​​of the distance error elements and the relative velocity error values ​​of the velocity error elements of the fitted error matrix are obtained.

[0093] Based on the line-of-sight information matrix, fitting error matrix, and pre-defined weighting matrix of the weighted least squares problem based on the recognition-removal strategy obtained by eliminating non-line-of-sight measurement information using the line-of-sight recognition algorithm based on Gaussian Bayes dynamic training, this algorithm is used.

[0094] Specifically, based on the observed and calculated distance values ​​between the moving target and the base station, and the observed and calculated relative velocity values, the distance error values ​​of the distance error elements and the relative velocity error values ​​of the velocity error elements of the fitted error matrix are obtained, including:

[0095] The distance is calculated according to the first preset formula, and the relative velocity is calculated according to the second preset formula. The expressions of the first preset formula and the second preset formula are as shown in expressions (1) and (2):

[0096]

[0097] in, Indicates the moving target is in The position to be estimated at a given time. Indicates the moving target is in The position to be estimated at a given time. This indicates the coordinates of the current line-of-sight base station. This indicates the period for synchronization of multiple base stations. This represents the calculated distance value. The calculated value representing relative velocity;

[0098] The distance error value is obtained according to expression (3) and used as an element of the distance error in the fitting error matrix;

[0099]

[0100] in, Represents distance observation value, Indicates distance error Element;

[0101] The relative velocity error value is obtained according to expression (4) and used as an element of the relative velocity error in the fitting error matrix;

[0102]

[0103] in, Represents the observed relative velocity values. Indicates speed error Element;

[0104] Based on distance error and speed error , obtained the The time estimation error;

[0105] according to T The estimation error of +1 short-time position sequence is used to obtain the fitting error matrix. Its expression is shown in formula (5);

[0106]

[0107] in T This indicates the length of the short-time position sequence.

[0108] Specifically, the expression for the preset weighting matrix of the weighted least squares problem based on the identification-elimination strategy includes:

[0109]

[0110] in, , , , , , For the fitting error vector, For weighted matrices, This is the line-of-sight information matrix. Represents a diagonal matrix. It is a distance-weighted matrix. The relative velocity weighting matrix is ​​obtained by changing... and Adjusting distance and relative velocity observations can alter the positioning results. It is a line-of-sight information matrix, and has .

[0111] In this invention, compared with the existing weighted least squares problem, this invention adopts a weighted least squares problem based on the identification-elimination strategy, and adds a disparity information matrix L to eliminate non-disparity information. Each matrix includes 2T+1 vectors. The first vector of each matrix contains only position information, and from the second vector onwards, it contains both relative velocity information and distance information.

[0112] It should be noted that, theoretically, the maximum likelihood estimation method can be used to... Simultaneous estimation is performed, and its expression includes:

[0113]

[0114] The function of this formula is to find a set of... This maximizes the probability under the obtained observation conditions, and is used to estimate the value of each position in the position sequence frame for theoretical calculation. However, the weighted least squares problem based on the identification-elimination strategy adopted in this invention is more feasible in practical engineering.

[0115] In this invention, the traditional TOA positioning method is used, and the position estimation expression at time k can be expressed as:

[0116]

[0117] From the The position estimation expression at time can be seen from this. The estimation accuracy is affected The direct impact of the estimation results means that relative motion velocity affects the positioning results through time integration. These two types of error factors will inevitably cause the accumulation of positioning errors, which in turn will cause the positioning system to diverge and fail.

[0118] Therefore, velocity accumulation error can be mitigated by simultaneously estimating multiple positions in a short-time position sequence. The initial position value is also included as an unknown variable in the estimation, improving positioning accuracy and stability while reducing its impact on the positioning result. This short-time sequence is referred to as a "short-time position sequence frame," and the position before the frame is used. The distance information is used as the starting point of the entire short time sequence frame, and its function is to disconnect the time "integration chain" by no longer introducing velocity information.

[0119] In one possible implementation, after determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the Levenberg-Marquardt algorithm iteration is terminated using the fitting residual and a preset maximum short-time position sequence frame length as the termination condition. During the iteration process, the sequence frame length is dynamically adjusted to obtain the optimal sequence frame length, including:

[0120] Based on the line-of-sight information matrix, it is determined that there are at least two base stations closest to the moving target;

[0121] The short-time position sequence frame length values ​​in the acquired sequence frame length set, arranged in ascending order;

[0122] Based on the comparison results between the short-time position sequence frame length value obtained in each iteration and the current position time value to be estimated, it is determined whether to reduce the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation.

[0123] Based on the first comparison result of the fitting residual values ​​of two adjacent fitting error matrices and the second comparison result of the actual short-time position sequence frame length value and the maximum short-time position sequence frame length value, determine whether to output the position estimation result of the position sequence frame at the time to be estimated.

[0124] Specifically, based on the comparison results of the short-time position sequence frame length value obtained in each iteration with the current position time value to be estimated, it is determined whether to reduce the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation, including:

[0125] In each iteration, if the obtained short-time position sequence frame length value is greater than or equal to the estimated position time value, then the short-time position sequence frame length value calculated in each iteration is reduced by 1 and assigned to the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation.

[0126] If the short-time position sequence frame length value obtained in each iteration is less than the time value of the position to be estimated, then the obtained short-time position sequence frame length value is used as the actual short-time position sequence frame length value used in the iterative calculation of the Levenberg-Marquardt algorithm.

[0127] In one embodiment provided by the present invention, such as Figure 4 As shown, Figure 4 The algorithm flowchart provided by this invention, starting from the beginning of the short-time sequence frame length iteration process, first resets the iteration process, letting... and assign values ,like Therefore, the first step is to determine the frame length of the short time sequence. To impose restrictions, that is Then load the line-of-sight information matrix Weighted matrix And the initial values ​​for the LM algorithm. ,in, Furthermore, this paper solves a high-dimensional nonlinear least squares problem under occlusion conditions, which uses a recognition-rejection strategy to suppress the impact of unreliable distance and relative velocity measurements on the system estimation accuracy. The estimated value Calculate the fitting residuals ,if and Then let Then proceed to the next loop until the fitted residual is found. Increase or sequence frame length If the value exceeds the preset range, finally, output a short-time position sequence. The estimated value In the estimation of short-time location sequence frames, if the length T is too large, the computational load will increase and the estimation accuracy will decrease.

[0128] Specifically, based on the line-of-sight information matrix, it is determined that there are at least two base stations closest to the moving target, including:

[0129] Obtain the current line-of-sight information matrix;

[0130] Then, determine whether the norm of the line-of-sight information matrix is ​​greater than 2. If it is less than 2, select the base station closest to the moving target from the obscured base stations, set the element corresponding to the base station in the line-of-sight information matrix to 1, and continue to determine whether the norm of the line-of-sight information matrix is ​​greater than 2. After it is greater than 2, it is determined that there are at least two base stations closest to the moving target.

[0131] It should be noted that when the measurement is line-of-sight information, It can be seen that in the first Time for shared One that needs to be estimated, its The space dimension is represented by , while The dimension is , Let represent the number of base stations. To ensure least-squares parity, the following relationship exists:

[0132]

[0133] for In a two-dimensional positioning scenario, if it is required that... If the positioning is based on real-time analysis, then the minimum base station deployment density required is 2, which is less than the 3 base stations commonly used for positioning. This results in higher positioning accuracy.

[0134] The following simulation, using two-dimensional scene positioning as an example, describes the embodiments of the present invention.

[0135] exist Three base stations are deployed within the scene of m. The locations of the base stations are as follows: , and The target carrying a smart mobile terminal moves counterclockwise around obstacles within the scene. Its trajectory includes both straight-line and arc-shaped movements to increase the target's maneuverability. The target's speed is 1 m / s, and the maximum length of the sequence frames is set. , The speed of sound is m / s, for ease of simulation, it is assumed that the distance and relative velocity measurements both follow a normal distribution, with variances of respectively. and .

[0136] The numerical simulation considers indoor occlusion factors, especially when the measurement point is at line-of-sight with the base station. Otherwise, it is 0, thus forming a line-of-sight information matrix. , Figure 5 To compare the numerical simulation position estimation of the two methods, measurement noise , After removing occlusion information, the TOA algorithm based on line-of-sight information is estimated using the maximum likelihood estimation method, which can theoretically reach the Cramero lower bound.

[0137] Therefore, this invention uses it as a benchmark algorithm for comparison. As can be seen from the scatter plot, the positioning results proposed in this invention have better stability and higher accuracy.

[0138] Figure 6For the performance comparison of the numerical simulation results of two types of algorithms, based on the positioning error, its cumulative distribution function (CDF) is statistically analyzed. The method of the present invention has higher performance, with a positioning error having an 80% probability of being less than 36 cm and a 60% probability of being less than 20 cm.

[0139] Since the TOA positioning method is only related to the noise level of the distance measurement, the noise level of the fixed relative velocity measurement is thus fixed, and the performance of the two types of methods is compared. The set in the simulation is a relatively high noise level compared to the target moving speed of 1 m / s. and are respectively set, and their performance comparison is as Figure 7 shown. It can be seen that the method proposed in the present invention has higher positioning accuracy and stability.

[0140] As Figure 8 shown, when and , after removing the occluder in the Figure 5 simulation scenario, the performance comparison in the case of 2 base stations and 3 base stations. Whether in the occluded scenario or the line-of-sight scenario, the method proposed in the present invention has higher positioning accuracy and stability. The operation is carried out on a PC host with an i7-8700 CPU and 8G of memory based on Matlab 2020b. The average time consumption for a single calculation is about 4.3 ms. It has high practical application value.

[0141] Next, the indoor occluded positioning method based on distance and relative velocity measurement sound technology proposed in the present invention is further evaluated for its performance through experiments in an actual scenario.

[0142] The positioning method proposed in this paper is verified and evaluated for its performance through experiments in an actual scenario. The experimental site is selected in the main teaching building hall on the third floor of the north campus of Chang'an University. The experimental scenario is an indoor occluded scenario. Figure 9 is the experimental scenario diagram, and the geometric size of the indoor positioning space is m. There are 4 columns in the hall as occluders to simulate the occluded positioning scenario. The broadcasting nodes and recording nodes of the experimental equipment achieve high-precision clock synchronization through the synchronization node based on LoRa technology. The audio chip is WM8978, and both the speaker and the microphone are consumer-grade MEMS components. The total cost of the sound components is less than 40 yuan.

[0143] In this experiment, the acoustic signal used for distance and relative velocity measurement estimation was a composite hyperbolic frequency modulation (HFM) signal with a sampling rate of 48000 Hz, a time-domain bandwidth of 0.05 s, a prefix and suffix time-domain bandwidth of 10 ms to suppress frequency leakage, and the lowest and highest frequencies of the falling signal component were 16555 Hz and 18555 Hz, respectively, while the corresponding frequencies of the rising signal component were 19555 Hz and 21555 Hz. The position update frequency of the positioning system was set to 1 Hz, meaning that a positioning experiment was performed on the target every 1 second.

[0144] like Figure 10 As shown, three base stations (recording nodes) are deployed on tripods within the scene at a height of 1.6m. Two of the base stations are positioned on the opposite side of any obstruction to increase signal blockage. The coordinates of the base stations are as follows: , and The experimenter held a tag (broadcast node) at a height of approximately 1.5m. Each base station and tag were synchronized via LoRa. The experimenter walked clockwise around the field along a white line on the ground at a normal walking pace. Figure 9 As shown, the travel path is a rectangle, and the vertex coordinates are respectively... , , and There are four pillars blocking the travel route from base stations 2 and 3. The travel route is used as the true reference value for positioning, and the distance from the estimated target position to the travel route is used as the positioning deviation. This is used to evaluate the positioning performance of the proposed algorithm.

[0145] Set the maximum length of the sequence frame , The speed of sound is m / s, based on an occlusion recognition-removal strategy, the maximum likelihood TOA estimation is solved using parameter space search, and the localization result is as follows: Figure 10 As shown, the location estimation results obtained using the method proposed in this paper are as follows. Figure 10 As shown in the figure, the proposed method exhibits higher stability through comparison. Its CDF statistics for positioning error are as follows: Figure 11 As shown, the algorithm in this paper has a 90% probability of being less than 0.78m and a 60% probability of being less than 0.31m, thus achieving higher positioning accuracy.

[0146] In another aspect, the present invention also provides an indoor positioning device 200 for use in obstructed environments, such as... Figure 12 As shown, the device includes:

[0147] The acquisition module 201 is used to obtain the calculated distance and relative speed between the moving target and the base station based on the acoustic signal information of the moving target's location at a certain moment.

[0148] Processing module 202 is used to construct a fitting error matrix by using the difference between the observed and calculated distance values ​​between the moving target and the base station at that moment and the difference between the observed and calculated relative motion speed values, and the line-of-sight information matrix obtained by using a line-of-sight recognition algorithm based on Gaussian Bayes dynamic training to eliminate non-line-of-sight measurement information, as the parameter matrix of a weighted least squares problem based on the recognition-elimination strategy, to obtain the expression of the target problem.

[0149] Output module 203 is used to solve the target problem expression using the Levenberg-Marquardt algorithm to obtain the position estimation result of the position sequence frame at the time of the position to be estimated. The initial values ​​of the Levenberg-Marquardt algorithm include the initial position sequence frame to be estimated that only represents distance information, the short-time position sequence frame to be estimated that simultaneously represents distance information and relative velocity, and the short-time position sequence frame length set. After determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the iteration of the Levenberg-Marquardt algorithm is terminated by the fitting residual and the preset maximum frame length. During the iteration process, the length of the sequence frame is dynamically adjusted to obtain the optimal length of the sequence frame.

[0150] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0151] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0152] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. An indoor positioning method in an obstructed environment, characterized in that, include: Based on the acoustic signal information of the moving target's location at a certain moment, the calculated distance and relative speed between the moving target and the base station are obtained. The fitting error matrix, constructed using the differences between the observed and calculated distance values ​​between the moving target and the base station at that moment, and the observed and calculated relative motion speed values, together with the line-of-sight information matrix obtained by eliminating non-line-of-sight measurement information using a Gaussian Bayes dynamic training line-of-sight recognition algorithm, is used as the parameter matrix for the weighted least squares problem based on the recognition-elimination strategy, thus obtaining the expression for the target problem. The Levenberg-Marquardt algorithm is used to solve the objective problem expression and obtain the position estimation results of the position sequence frames at the time of the position to be estimated. The initial values ​​of the Levenberg-Marquardt algorithm include the initial position sequence frames to be estimated that only represent distance information, the short-time position sequence frames to be estimated that simultaneously represent distance information and relative velocity, and the short-time position sequence frame length set. After determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the iteration of the Levenberg-Marquardt algorithm is terminated by the fitting residual and the preset maximum frame length. During the iteration process, the length of the sequence frames is dynamically adjusted to obtain the optimal sequence frame length.

2. The indoor positioning method in an obstructed environment as described in claim 1, characterized in that, The step of obtaining the distance calculation value and relative motion speed calculation value between the moving target and the base station based on the acoustic signal information of the moving target's position at a certain moment includes: Obtain a composite signal comprising two hyperbolic frequency modulated signals; The first and second delay information of the composite signal are obtained by two matched filters, respectively. Based on the first and second time delay information, the estimated time delay and the estimated Doppler factor of the composite signal are obtained. The distance between the moving target and the base station and the relative speed are calculated by multiplying the estimated time delay and the estimated Doppler factor with the speed of sound, respectively.

3. The indoor positioning method in an obstructed environment as described in claim 2, characterized in that, The fitting error matrix constructed using the differences between the observed and calculated distance values ​​between the moving target and the base station at that moment, and the observed and calculated relative motion speed values, together with the line-of-sight information matrix obtained by eliminating non-line-of-sight measurement information using a Gaussian Bayes dynamic training line-of-sight recognition algorithm, serves as the parameter matrix for the weighted least squares problem based on the recognition-elimination strategy, including: Using the relative motion velocity observation value and distance observation value between the moving target and the current line-of-sight base station at any given time, the coordinates of the current base station, and the position to be estimated of the moving target at the current time as parameters, the distance calculation value is obtained by using the first preset formula, and the relative velocity observation value is obtained by using the second preset formula. Based on the observed and calculated distance values ​​between the moving target and the base station, as well as the observed and calculated relative velocity values, the distance error values ​​of the distance error elements and the relative velocity error values ​​of the velocity error elements of the fitted error matrix are obtained. Based on the line-of-sight information matrix, fitting error matrix, and pre-defined weighting matrix of the weighted least squares problem based on the recognition-removal strategy obtained by eliminating non-line-of-sight measurement information using the line-of-sight recognition algorithm trained by Gauss-Bayes dynamic training, this algorithm is used.

4. The indoor positioning method in an obstructed environment as described in claim 3, characterized in that, The step of obtaining the distance error value of the distance error element and the relative velocity error value of the velocity error element of the fitted error matrix based on the observed and calculated distance values ​​between the moving target and the base station, and the observed and calculated relative velocity values, includes: The distance is calculated according to the first preset formula, and the relative velocity is calculated according to the second preset formula. The expressions of the first preset formula and the second preset formula are as shown in expressions (1) and (2): in, Indicates the moving target is in The position to be estimated at a given time. Indicates the moving target is in The position to be estimated at a given time. This indicates the coordinates of the current line-of-sight base station. This indicates the period for synchronization of multiple base stations. This represents the calculated distance value. The calculated value representing relative velocity; The distance error value is obtained according to expression (3) and used as an element of the distance error in the fitting error matrix; in, Represents distance observation value, Indicates distance error Element; The relative velocity error value is obtained according to expression (4) and used as an element of the relative velocity error in the fitting error matrix; in, Represents the observed relative velocity values. Indicates speed error Element; Based on distance error and speed error , obtained the The time estimation error; according to T The estimation error of +1 short-time position sequence is used to obtain the fitting error matrix. Its expression is shown in formula (5); in T This indicates the length of the short-time position sequence.

5. The indoor positioning method in an obstructed environment as described in claim 4, characterized in that, The expression for the pre-defined weighting matrix of the weighted least squares problem based on the identification-elimination strategy includes: in, , , , , , For the fitting error vector, For weighted matrices, This is the line-of-sight information matrix. Represents a diagonal matrix. It is a distance-weighted matrix. The relative velocity weighting matrix is ​​obtained by changing... and Adjusting distance and relative velocity observations can alter the positioning results. It is a line-of-sight information matrix, and has .

6. The indoor positioning method in an obstructed environment as described in claim 1, characterized in that, After determining, based on the line-of-sight information matrix, that there are at least two base stations closest to the moving target, the Levenberg-Marquardt algorithm iteration is terminated using the fitting residual and a preset maximum short-time position sequence frame length as the terminating condition. During the iteration process, the sequence frame length is dynamically adjusted to obtain the optimal sequence frame length, including: Based on the line-of-sight information matrix, it is determined that there are at least two base stations closest to the moving target; The short-time position sequence frame length values ​​in the acquired sequence frame length set, arranged in ascending order; Based on the comparison results between the short-time position sequence frame length value obtained in each iteration and the current position time value to be estimated, it is determined whether to reduce the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation. Based on the first comparison result of the fitting residual values ​​of two adjacent fitting error matrices and the second comparison result of the actual short-time position sequence frame length value and the maximum short-time position sequence frame length value, determine whether to output the position estimation result of the position sequence frame at the time to be estimated.

7. The indoor positioning method in an obstructed environment as described in claim 6, characterized in that, The step of determining whether to reduce the actual short-time position sequence frame length used in the Levenberg-Marquardt algorithm iteration calculation based on the comparison results of the short-time position sequence frame length value obtained in each iteration with the current position time value to be estimated includes: In each iteration, if the obtained short-time position sequence frame length value is greater than or equal to the estimated position time value, then the short-time position sequence frame length value calculated in each iteration is reduced by 1 and assigned to the short-time position sequence frame length value actually used in the Levenberg-Marquardt algorithm iteration calculation. If the short-time position sequence frame length value obtained in each iteration is less than the time value of the position to be estimated, then the obtained short-time position sequence frame length value is used as the actual short-time position sequence frame length value used in the iterative calculation of the Levenberg-Marquardt algorithm.

8. The indoor positioning method in an obstructed environment as described in claim 6, characterized in that, The step of determining, based on the line-of-sight information matrix, that there are at least two base stations closest to the moving target includes: Obtain the current line-of-sight information matrix; Then, determine whether the norm of the line-of-sight information matrix is ​​greater than 2. If it is less than 2, select the base station closest to the moving target from the obscured base stations, set the element corresponding to the base station in the line-of-sight information matrix to 1, and continue to determine whether the norm of the line-of-sight information matrix is ​​greater than 2. After it is greater than 2, it is determined that there are at least two base stations closest to the moving target.

9. An indoor positioning device for use in obstructed environments, characterized in that, include: The acquisition module is used to obtain the calculated distance and relative speed between the moving target and the base station based on the acoustic signal information of the moving target's location at a certain moment. The processing module is used to construct a fitting error matrix by using the difference between the observed and calculated distance values ​​between the moving target and the base station at that moment, and the difference between the observed and calculated relative motion speed values, and the line-of-sight information matrix obtained by using a line-of-sight recognition algorithm based on Gaussian Bayes dynamic training to eliminate non-line-of-sight measurement information, as the parameter matrix of the weighted least squares problem based on the recognition-elimination strategy, to obtain the expression of the target problem. The output module is used to solve the objective problem expression using the Levenberg-Marquardt algorithm to obtain the position estimation results of the position sequence frames at the time of the position to be estimated. The initial values ​​of the Levenberg-Marquardt algorithm include the initial position sequence frames to be estimated that only represent distance information, the short-time position sequence frames to be estimated that simultaneously represent distance information and relative velocity, and the short-time position sequence frame length set. After determining that there are at least two base stations closest to the moving target based on the line-of-sight information matrix, the iteration of the Levenberg-Marquardt algorithm is terminated by the fitting residual and the preset maximum frame length. During the iteration process, the length of the sequence frames is dynamically adjusted to obtain the optimal length of the sequence frames.

Images