IMU (Inertial Measurement Unit) fusion positioning method based on double-weighted neighbor matching

Abstract

The invention relates to an IMU (Inertial Measurement Unit) fusion positioning method based on double-weighted neighbor matching, which comprises the following steps of: firstly, dividing a positioning area into grids in a layered manner, determining a reference point, and collecting all signal strength (RSSI) data of WiFi (Wireless Fidelity) which can be received within 10 seconds at the reference point; then, establishing a feature enhanced fingerprint database based on an AP optimization strategy of an optimal reception rate; in the real-time positioning stage, a fingerprint database with enhanced features is matched based on a double-weighted neighbor algorithm (DWKNN), and a target position information observation value is obtained; meanwhile, an inertial measurement unit (IMU) is used for collecting inertial data, and a position estimation value is output through a pedestrian track plotting algorithm; and finally, through an extended Kalman filter (EKF) model, fusing the position observation value matched by the WiFi and the position estimation value calculated by the IMU to obtain an optimal position estimation value. The method can effectively reduce the influence of the abnormal value of the weak point in the complex indoor scene on the positioning effect, further improves the positioning precision, and overcomes the limitation caused by a single technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of navigation and positioning, and relates to an IMU indoor positioning method based on double-weighted nearest neighbor matching. Background Art

[0002] The statements in this part only provide background technical information related to the present disclosure, and these statements may constitute prior art. In the process of implementing the present invention, the inventors found that at least the following problems exist in the prior art.

[0003] The widespread presence of WiFi access points (APs) and the advantages of WiFi signals facilitate indoor positioning and can significantly save the construction time and cost of WiFi-based positioning systems. Small and low-power inertial sensors, such as acceleration sensors, gyroscopes, and pressure sensors, have become common components in the field of positioning and navigation, and personnel positioning is achieved through the pedestrian dead reckoning (PDR) algorithm. The positioning technology that combines WiFi and IMU (inertial measurement unit) overcomes the limitations brought by a single technology and has thus attracted increasing attention.

[0004] For example, the patent with the application number 202411335560.2 and the name "Indoor Positioning Method, System and Device Based on Fingerprint Database and Multi-Sensor Fusion" fuses a fingerprint database and multi-sensors. Its method includes: obtaining the positioning result of a target to be measured in a fingerprint database; obtaining sensor data of a sensor network, and calculating the forward information of the target to be measured based on the sensor data; calculating a heading angle based on the sensor data, and calculating the position information of the target to be measured based on the forward information and the heading angle; and determining the target positioning result of the target to be measured based on the positioning result and the position information. The purpose of this patent is to improve the accuracy of indoor positioning.

[0005] However, the applicant found that the above method did not achieve the claimed effect of improving positioning accuracy. Through in-depth research, it was found that the positioning error is mainly related to the WiFi indoor positioning method and the accuracy of inertial sensors. First, the WiFi indoor positioning method is easily interfered by environmental factors, resulting in signal strength fluctuations. Among them, the difference in AP quality is one of the core problems in WiFi positioning. The difference in AP quality reduces the effectiveness of the fingerprint database, thus significantly affecting the positioning effect. In addition, relying on existing wireless access points (APs) means that the quality of APs cannot be controlled, which leads to differences in WiFi signals and has a significant impact on positioning accuracy. Second, because the accuracy of inertial sensors is limited, the irregular shaking of the limbs during walking will also affect the accuracy of the data. Therefore, over time, the error in PDR positioning will gradually accumulate, resulting in an increase in long-term positioning error. Summary of the Invention

[0006] In view of the above problems, an object of the present invention is to solve a part of the problems in the prior art, or at least alleviate these problems.

[0007] An IMU fusion positioning method based on double weighted nearest neighbor matching includes the following steps:

[0008] Divide the positioning area into grids in layers and determine reference points, and collect RSSI (signal strength) data of all WiFis that can be received at the reference points for 10 s;

[0009] Based on all the collected RSSI data and the AP optimization strategy based on the optimal reception rate, establish a fingerprint database with enhanced features;

[0010] In the real-time positioning stage, match the fingerprint database with enhanced features based on the double weighted nearest neighbor algorithm (DWKNN) to obtain an observed value of the target position information;

[0011] Use an IMU to collect inertial data, and calculate an estimated position value through a pedestrian dead reckoning algorithm (PDR);

[0012] Through an extended Kalman filter (EKF) model, fuse the observed value of the position information matched by Wi-Fi and the estimated position value calculated by the IMU to obtain an optimal estimated position value.

[0013] Based on all the collected RSSI data and the AP optimization strategy based on the optimal reception rate, establish a fingerprint database with enhanced features, including the following steps:

[0014] Dynamically set a threshold for the signal strength according to the distribution of all the collected RSSI data, calculate the threshold using the RSSI mean and standard deviation, and further eliminate outliers generated during the collection process using statistical methods;

[0015] Define the AP reception ratio as:

[0016]

[0017] In the formula, m is the number of reference points that can receive this AP, n is the total number of reference points, and β is a stability coefficient;

[0018] Traverse all the reference point data and count the reception ratio of each AP, set the basic threshold of the reception ratio to 60%, eliminate the APs with a reception ratio lower than this value, and further eliminate the APs with large fluctuations in the coverage range. Finally, use them to construct a fingerprint database with enhanced features.

[0019] Match the fingerprint database with enhanced features based on the double weighted nearest neighbor algorithm (DWKNN) to obtain an observed value of the target position information, including the following steps:

[0020] Calculate the spatial distance and Euclidean distance from all reference points (RP) to the test point (TP), and select the smallest K reference points based on these two distances respectively to obtain two sets of reference points; calculate the spatial distance weights of the set of reference points selected by the spatial distance, and then calculate the Euclidean distance weights of the set of reference points selected by the Euclidean distance;

[0021] Screen out the common reference points from the two sets of reference points, and calculate the estimated position of the TP based on the common reference points and their weights. The formula is as follows:

[0022]

[0023] Use the normalized weighting method for fusion; among them, is the spatial distance weight of the K reference points with the smallest spatial distance; is the Euclidean distance weight of the K reference points with the smallest Euclidean distance, (x i , y i ) is the reference point coordinate.

[0024] Furthermore, select and calculate the spatial distance weights according to the spatial distance, including the following steps:

[0025] Calculate the spatial distance SD of the TP to the i-th reference point (RP i ), which can be obtained by the formula:

[0026]

[0027] where SD i is the spatial distance of the i-th RP; N is the number of APs that can be detected simultaneously at the TP and RP; d j is the distance from the TP to the j-th AP calculated by the signal attenuation model; is the distance from the i -th RP to the j-th AP; is the spatial distance from the i-th RP to the j-th AP.

[0028] can be obtained by the formula:

[0029]

[0030] Where η is the path loss exponent of the signal propagation in the real space; the values of η and P(d0) are fitted using the measured RSS data of the APs at known locations in the experimental area; the fitted values of η and P(d0) are 3.3 and -36 dBm respectively; d is the distance between the receiver and the transmitter; P(d) is the RSS value received by the receiver at a distance d from the transmitter, and P(d0) is the RSS value received at a distance d0, where d0 is the reference distance;

[0031] The distance from the receiver to the transmitter is calculated by the following formula:

[0032]

[0033] Similarly, the formula for calculating d j can be obtained:

[0034]

[0035] where RSS is the signal strength value at a distance d0 from the j-th AP; RSS j is the signal strength value received at the TP location from the j-th AP;

[0036] Similarly, as the formula:

[0037]

[0038] where, is the signal strength value received at the i-th RP location from the j-th AP;

[0039] Select K reference points with the smallest distances to calculate the weights; the weight formula for the spatial distance is as follows:

[0040]

[0041] Furthermore, the steps for selecting and calculating the Euclidean distance weights according to the Euclidean distance are as follows:

[0042] Calculate the Euclidean distance of the TP to the i-th reference point (RP i ), and the calculation formula is:

[0043]

[0044] Select K reference points with the smallest Euclidean distances for calculating the weights of the Euclidean distance, and the calculation formula is as follows:

[0045]

[0046] The pedestrian dead reckoning (PDR) algorithm detects the pedestrian's step frequency, estimates the step length, and calculates the heading through an accelerometer and a gyroscope, and finally obtains the walking trajectory by superimposing displacements; the coordinate position of the pedestrian is calculated by superimposing the following formula:

[0047]

[0048] In the formula, k represents the k-th moment, (x k , y k ) respectively represent the coordinate position at the k-th moment, (x k-1 , y k-1 ) respectively represent the coordinate position at the (k - 1)-th moment, S k represents the step length at the k-th moment, and θ k represents the heading at the k-th moment.

[0049] By using the extended Kalman filter (EKF) model, fusing the observed value of the WiFi-matched position information and the estimated value of the position deduced by the IMU, the optimal position estimate value is obtained, including the following steps:

[0050] Establish a fusion state model and an observation model:

[0051]

[0052] In the model, x k and y k represent the x-axis coordinate and y-axis coordinate of the pedestrian at the k-th step; d k represents the step length at the k-th step; θ k is the heading at the k-th step; Δd k is the step length increment at the k-th moment; represents the result of WiFi positioning, and W k , V k are the system noise and the observation noise respectively;

[0053] Perform prediction and update through the following algorithm to obtain the optimized positioning result, that is, the optimal position estimate value;

[0054]

[0055] Among them is the prior estimate of the measurement, is the prior estimate of the state, and H k is the observation matrix; is the prior estimate of the error covariance, Q is the error variance matrix, and F k is the Jacobian matrix; is the transpose of the F k matrix;

[0056]

[0057] Where K k is the Kalman filter gain at the k-th step, and R k is the covariance matrix of the measurement noise at the k-th step. X k is the estimated value at the k-th step, and P k is the posterior error covariance matrix at the k-th step. is the transpose of H k ; Z k is the measurement matrix at the k-th step; I is the identity matrix.

[0058] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, the steps of the IMU fusion positioning method based on double-weighted nearest neighbor matching are implemented.

[0059] The present invention has the following beneficial effects:

[0060] The AP optimization strategy based on the optimal reception rate proposed by the present invention, and establishing a fingerprint database with enhanced features is beneficial to reducing the influence of outliers at weak points in complex indoor scenarios on the positioning effect; and in combination with the double-weighted nearest neighbor algorithm matching algorithm (DWKNN) proposed by the present invention, it can improve the single-weight strategy of the existing WKNN algorithm, combine the weights of spatial distance and Euclidean distance, and further improve the positioning accuracy; finally, by fusing the WiFi positioning and PDR positioning results through an extended Kalman filter model, the limitations brought by a single technology are overcome. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a flowchart of an IMU indoor positioning method based on double-weighted nearest neighbor matching in a preferred embodiment of the present invention;

[0062] Figure 2 is a block diagram of an AP screening algorithm in a preferred embodiment of the present invention.

[0063] Figure 3 is a block diagram of the DWKNN algorithm process in a preferred embodiment of the present invention.

[0064] Figure 4 is a positioning trajectory diagram of different algorithms in a preferred embodiment of the present invention.

[0065] Figure 5 is an error accumulation distribution diagram of different algorithms in a preferred embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The following further describes the present invention with reference to the accompanying drawings. The embodiments of the present invention are only used to illustrate the present invention and do not limit the present invention. Without departing from the technical idea of the present invention, various substitutions and changes made according to common general technical knowledge and conventional means in the art should be included within the scope of the present invention.

[0067] An IMU indoor positioning method based on double weighted nearest neighbor matching provided by the present invention, as shown in the appendix Figure 1 is as follows:

[0068] Step a: Divide the positioning area into layers and grids and determine reference points, and collect RSSI data of all WiFis that can be received at the reference points for 10s.

[0069] Step b: Based on all the collected RSSI data and the AP optimization strategy based on the optimal reception rate, establish a fingerprint database with enhanced features.

[0070] The AP optimization strategy with the optimal reception rate, the specific process of which is as follows:

[0071] First, dynamically set the threshold of the signal strength according to the distribution of all the collected RSSI data, calculate the threshold using the RSSI mean and standard deviation, and further eliminate the outliers generated during the collection process by statistical methods to improve the quality of the RSSI data.

[0072] Define the AP reception ratio as:

[0073]

[0074] where m is the number of reference points that can receive this AP, n is the total number of reference points, and β is the stability coefficient.

[0075] Traverse all the reference point data and count the reception ratio of each AP. Set the basic threshold of the reception ratio to 60%, eliminate the APs with a reception ratio lower than this value, and further consider the distribution uniformity of the reception ratio. Calculate the standard deviation of the reception ratio and eliminate the APs with large fluctuations in the coverage range. Finally, use them to construct the fingerprint database. The flowchart of the AP screening algorithm is as shown in the appendix Figure 2 as follows.

[0076] Step c: In the real-time positioning stage, match the fingerprint database with enhanced features based on the double weighted nearest neighbor algorithm (DWKNN) to obtain the observed value of the target position information.

[0077] The indoor positioning method based on the double weighted nearest neighbor algorithm (DWKNN) involved in step c improves the single weight strategy of the existing WKNN algorithm by combining the weights of the spatial distance and the Euclidean distance. The flowchart of the algorithm is as shown in the appendix Figure 3 as follows.

[0078] First, calculate the spatial distance and Euclidean distance from all reference points (RP) to the test point (TP), and select the K nearest reference points based on these two distances respectively. Among them, the spatial distance is calculated through the signal attenuation model, which reflects the actual physical distance of signal propagation; the Euclidean distance is calculated based on geometric relationships, which reflects the straight-line distance between the reference point and the test point. The core difference between the two lies in their relationship with the received signal strength (RSS): the Euclidean distance has a linear relationship with the RSS value, while the spatial distance has an exponential relationship with the RSS value.

[0079] The spatial distance SD of the TP to the i-th reference point (RP i ) can be obtained by the formula:

[0080]

[0081] Where SD i is the spatial distance of the i-th RP; N is the number of APs that can be detected simultaneously at the TP and RP; the value of m may be different among different RPs. d j is the distance from the TP to the j-th AP calculated by the signal attenuation model; is the distance from the i -th RP to the j-th AP; is the spatial distance from the i-th RP to the j-th AP. It can be obtained by the formula:

[0082]

[0083] In the formula, η is the path loss exponent of signal propagation in the real space. In this article, the values of η and P(d0) are fitted with the measured RSS data of APs at known positions in the experimental area; the fitted values of η and P(d0) are 3.3 and -36 dBm respectively; d is the distance between the receiver and the transmitter; P(d) is the RSS value received by the receiver at a distance d from the transmitter, and P(d0) is the RSS value received at a distance d0, where d0 is the distance reference.

[0084] The signal attenuation model shows the relationship between the RSS received by the receiver and the distance between the receiver and the transmitter; the distance between the receiver and the transmitter can be calculated by the following formula:

[0085]

[0086] Similarly, the formula for calculating d j can be obtained:

[0087]

[0088] Where RSS is the signal strength value at a distance d0 from the j-th AP; RSSj It is the signal strength value received from the j-th AP at the TP position.

[0089] Similarly, As the formula:

[0090]

[0091] Wherein, It is the signal strength value received from the j-th AP at the i-th RP position.

[0092] Then, calculate the spatial distance weight and Euclidean distance weight of each RP. The weight value is inversely proportional to the distance. Select the K RPs with the smallest distances to calculate the weights. The formula for the spatial distance weight is as follows:

[0093]

[0094] The Euclidean distance ED is used to represent the similarity between the measurement test point TP and the reference point RP. The calculation formula is:

[0095]

[0096] After calculating the Euclidean distances of all reference points, select the smallest K reference points to calculate the weights of the Euclidean distances. The calculation formula is as follows:

[0097]

[0098] Next, screen out the common RPs from the two groups of RPs. These common RPs are used to calculate the coordinates of the test point (TP). By fusing the two weights, the prediction or positioning accuracy is improved. This patent selects the normalized weighting method. The normalized weighting method combines the operations of weighting and normalization, and is commonly used to ensure that the values of different weights are within a unified range [0,1]. This method is usually used for multi-dimensional fusion and multi-index evaluation.

[0099] The formula is as follows:

[0100]

[0101] Finally, based on these common RPs and their weights, calculate the estimated position of TP. The formula is as follows:

[0102]

[0103] Step d: Meanwhile, use the IMU to collect inertial data, and estimate the position value through the Pedestrian Dead Reckoning (PDR) algorithm.

[0104] In step d, the PDR algorithm detects the pedestrian's step frequency, estimates the step length, and calculates the heading through the accelerometer and gyroscope. Finally, the walking trajectory is obtained by superimposing displacements. The coordinate position of the pedestrian can be calculated by superimposing the following formula:

[0105]

[0106] In the formula, k represents the k-th moment, (x k , y k ) respectively represent the coordinate position at the k-th moment, (x k-1 , y k-1 ) respectively represent the coordinate position at the (k - 1)-th moment, S k represents the step length at the k-th moment, and θ k represents the heading at the k-th moment.

[0107] Step e: Through the Extended Kalman Filter (EKF) model, fuse the position information observation value of WiFi matching and the position estimation value calculated by IMU to obtain the optimal position estimation value.

[0108] In step e, the WiFi / PDR fusion algorithm based on EKF is as shown in the appendix Figure 4 . Since the WiFi fingerprint positioning obtains the absolute position and the PDR algorithm obtains the cumulative value of the step length and heading angle, a fusion state model and an observation model are established:

[0109]

[0110] In the model, x k and y k represent the x-axis coordinate and y-axis coordinate of the pedestrian at the k-th step; d k represents the step length at the k-th step; θ k is the heading at the k-th step; Δd k is the step length increment at the k-th moment; represents the result of WiFi positioning, and W k , V k are the system noise and observation noise respectively.

[0111] The entire filtering algorithm is divided into two stages: prediction and update. Prediction requires advancing the current state forward and estimating the error covariance to obtain the prior estimate for the next step.

[0112]

[0113] Among them is the prior estimate of the measurement, is the prior estimate of the state, and H k is the identity matrix; is the prior estimate of the error covariance, Q is the error variance matrix, and F kis the Jacobian matrix; is F k the transpose of the matrix.

[0114]

[0115] where K k is the Kalman filter gain at the k-th step, and R k is the covariance matrix of the measurement noise at the k-th step. X k is the estimated value at the k-th step, and P k is the posterior error covariance matrix at the k-th step. is the transpose of H k ; Z k is the measurement matrix at the k-th step; I is the identity matrix. Through the above filtering algorithm, an optimized positioning result can be obtained.

[0116] Furthermore, in order to verify the feasibility and accuracy of the method of the present invention in indoor positioning, an indoor positioning experiment was conducted. Table 1 gives the main technical indicators of the used IMU, and the WiFi module model is ESP32.

[0117] Table 1

[0118]

[0119] The experimental location is the laboratory office. Figure 4 For the comparison chart of the positioning effects of the improved single-weight DKWNN algorithm, PDR algorithm and EKF fusion algorithm, it can be clearly seen that the positioning accuracy of the method of the present invention is higher.

[0120] Figure 5 For the error accumulation distribution chart of the three positioning algorithms, the probability that the positioning accuracy of the EKF fusion algorithm is less than 1.53 m reaches 96.3%. The EKF fusion positioning algorithm is superior to the single positioning and can effectively improve the accuracy. The specific indicators are shown in Table 1.

[0121] The data in Table 2 shows the performance comparison of four positioning algorithms (WKNN, DWKNN, PDR, and EKF fusion) under different evaluation indicators, including root mean square error (RMSE), mean error, and loop error. From various error data, the EKF fusion algorithm performs the best and is significantly lower than the other three algorithms. It shows that the EKF fusion algorithm of the present invention has obvious advantages in positioning accuracy and can estimate the position more accurately.

[0122] Table 2

[0123]

[0124] A computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the IMU fusion positioning method based on double weighted nearest neighbor matching are implemented.

[0125] The parts not detailedly disclosed in the present invention belong to the well-known technologies in the art. Although the illustrative specific embodiments of the present invention are described above for the convenience of those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions made using the concept of the present invention are within the scope of protection.

Claims

1. The IMU fusion positioning method based on double weighted nearest neighbor matching is characterized by: The following steps are involved: Divide the positioning area into grids and determine the reference points. Collect the RSSI (signal strength) data of all WiFi that can be received within 10 seconds at the reference points. According to all collected RSSI data, a fingerprint library with enhanced features is established based on the AP optimization strategy with the best reception rate; In the real-time positioning stage, the fingerprint library with feature enhancement is matched based on the double weighted nearest neighbor algorithm (DWKNN) to obtain the target location information observation value; Use IMU to collect inertial data and calculate position estimate through pedestrian dead reckoning (PDR); The extended Kalman filter (EKF) model is used to fuse the location information observations matched by WiFi and the location estimation values ​​calculated by IMU to obtain the optimal location estimation value.

2. The IMU fusion positioning method based on double-weighted nearest neighbor matching according to claim 1, characterized in that: According to all collected RSSI data, based on the AP optimization strategy with the best reception rate, a fingerprint library with feature enhancement is established, including the following steps: Dynamically set the signal strength threshold based on the distribution of all collected RSSI data, calculate the threshold using the RSSI mean and standard deviation, and use statistical methods to further eliminate outliers generated during the collection process; Define the AP receiving ratio as: Where m is the number of reference points that can receive the AP, n is the total number of reference points, and β is the stability coefficient; Traverse all reference point data and count the reception ratio of each AP. Set the basic threshold of the reception ratio to 60%, remove APs with reception ratios lower than this value, and further remove APs with large coverage fluctuations. Finally, the fingerprint library with feature enhancement is constructed.

3. The IMU fusion positioning method based on double-weighted nearest neighbor matching according to claim 1, characterized in that: Based on the double weighted nearest neighbor algorithm (DWKNN), the fingerprint library with feature enhancement is matched to obtain the target location information observation value, including the following steps: Calculate the spatial distance and Euclidean distance from all reference points (Reference Point, RP) to the test point (Test Point, TP), and select the smallest K reference points based on these two distances to obtain two groups of reference points; calculate the spatial distance weight of a group of reference points selected by the spatial distance, and then calculate the Euclidean distance weight of a group of reference points selected according to the Euclidean distance; A common reference point is selected from the two groups of reference points, and the estimated position of the TP is calculated based on the common reference point and its weight, as shown in the following formula: The normalized weighted method is used for fusion; is the spatial distance weight of the K reference points with the smallest spatial distance; is the Euclidean distance weight of the K reference points with the smallest Euclidean distance, (x i ,y i ) are the reference point coordinates.

4. The IMU fusion positioning method based on double-weighted nearest neighbor matching according to claim 3 is characterized in that: Selecting and calculating the spatial distance weight according to the spatial distance includes the following steps: Calculate TP for the i-th reference point (RP i ) can be obtained by the formula: Among them SD i is the spatial distance of the i-th RP; N is the number of APs that can be detected simultaneously at TP and RP; d j The distance from TP to the jth AP calculated by the signal attenuation model; For the i The distance from the jth RP to the jth AP; is the spatial distance from the i-th RP to the j-th AP. It can be obtained by the formula: Where η is the path loss exponent of the signal propagating in real space; the values ​​of η and P(d0) are fitted with the measured RSS data of the AP at a known position in the experimental area; the fitted values ​​of η and P(d0) are 3.3 and -36 dBm respectively; d is the distance between the receiver and the transmitter; P(d) is the RSS value received by the receiver at a distance d from the transmitter, and P(d0) is the RSS value received at a distance d0, where d0 is the distance reference; The distance from the receiver to the transmitter is calculated using the following formula: Similarly, we can calculate d j The formula is: Where RSS is the signal strength value at d0 from the jth AP; RSS j is the signal strength value received from the jth AP at the TP location; Similarly, Such as the formula: in, is the signal strength value received from the jth AP at the i-th RP position; Select K with the smallest distances to calculate the weight; the weight formula of the spatial distance is as follows:

5. The IMU fusion positioning method based on double-weighted nearest neighbor matching according to claim 3 is characterized in that: The steps to select and calculate the Euclidean distance weight according to the Euclidean distance are as follows: Calculate TP for the i-th reference point (RP i ) is calculated as: Select K reference points with the smallest Euclidean distance to calculate the weight of the Euclidean distance. The calculation formula is as follows:

6. The IMU fusion positioning method based on double-weighted nearest neighbor matching according to claim 1, characterized in that: The Pedestrian Dead Reckoning (PDR) algorithm uses an accelerometer and a gyroscope to detect the pedestrian's cadence, estimate the step length, and calculate the heading. Finally, the walking trajectory is obtained by superimposing the displacement. The coordinate position of the pedestrian is calculated by superimposing the following formula: In the formula, k represents the kth moment, (x k ,y k ) represent the coordinate position at time k, (x k-1 ,y k-1 ) represent the coordinate position at time k-1, S k represents the step size at time k, θ k represents the heading at time k.

7. The IMU fusion positioning method based on double-weighted nearest neighbor matching according to claim 1, characterized in that: The extended Kalman filter (EKF) model is used to fuse the location information observation value matched by WiFi and the location estimate value calculated by IMU to obtain the optimal location estimate, including the following steps: Establish fusion state model and observation model: Model x k and k represents the x-axis coordinate and y-axis coordinate of the kth pedestrian; d k represents the step length of the kth step; θ k is the heading of the kth step; Δd k is the step size increment at time k; Indicates the positioning result by WiFi, W k 、V k It can be divided into system noise and observation noise; The following algorithm is used to predict and update to obtain the optimized positioning result, that is, the optimal estimated value of the position; in is the measurement prior estimate, is the state prior estimate, H k is the observation matrix; is the prior estimate of the error covariance, Q is the error variance matrix, and F k is the Jacobian matrix; Yes F k Transpose of a matrix; Where K k is the Kalman filter gain of the kth step, R k is the covariance matrix of the k-th step measurement noise. k is the estimated value at step k, P k is the posterior error covariance matrix of the k-th step, H k The transpose of Z k is the measurement matrix of the kth step; I is the unit matrix.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the IMU fusion positioning method based on double-weighted nearest neighbor matching described in any one of claims 1 to 7 are implemented.

Images