An Indoor Pedestrian Localization Method Combining MEMS-IMU and Wi-Fi

Abstract

This invention relates to an indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi. It employs a position recursion model established using a pedestrian dead reckoning algorithm, fundamentally mitigating the impact of severe system error drift from low-cost MEMS-IMUs on the free inertial navigation solution. Simultaneously, by establishing gyroscope-based and Wi-Fi-based measurement models, the observed values ​​directly and independently participate in the measurement update of the extended Kalman filter algorithm, thus avoiding the heavy dependence of the inertial navigation mechanism on the IMU output in traditional integrated strategies. Furthermore, based on the selection of IMU measurement values, the system error model established for the gyroscope in the IMU significantly reduces the impact of the original IMU output noise, as both the original data and system error participate in the extended Kalman filter update.

Description

Technical Field

[0001] This invention relates to the field of indoor positioning and navigation technology, and more specifically, to an indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi. Background Technology

[0002] In modern cities, people spend approximately 80% of their time indoors. In the mobile internet era, 70% of mobile phone-based communication, mobile consumption, payments, and data connections, as well as 80% of internet data interaction, occur indoors. This makes indoor positioning increasingly widely applicable and in demand, and the market value related to indoor positioning and location services is also growing. Developing effective indoor pedestrian navigation and positioning technology can significantly reduce users' self-localization time, providing great convenience for users' activities and work.

[0003] The primary task of indoor pedestrian navigation and positioning technology at this stage is to reduce system operating costs and improve system portability while ensuring system positioning accuracy. In modern society, smartphones are essential for pedestrians and are almost always in their hands. Smartphones often embed Micro-Electro-Mechanical Systems-Inertial Measurement Units (MEMS-IMUs), which can be used to achieve fully autonomous inertial navigation and positioning. Compared to other navigation methods, inertial navigation algorithms do not require any external equipment, are not limited by external environmental conditions, and have greater freedom of movement. In 2020, Deng et al., in their paper "An Indoor Positioning Algorithm Based on Accurate Heading Angle Correction," used an improved heuristic heading angle correction algorithm to accurately correct heading angles in complex indoor environments, thereby improving the accuracy of pedestrian dead reckoning and positioning. However, the extremely low cost of miniature inertial measurement units (gyroscopes, accelerometers, and magnetometers, etc.) significantly reduces their performance, making them unsuitable for standalone indoor pedestrian navigation. Therefore, MEMS-IMU-based inertial navigation systems require information fusion with other positioning technologies to achieve reliable positioning. Indoor environments are often equipped with Wireless Local Area Networks (WLANs), enabling Wireless Fidelity (Wi-Fi), and public places currently provide free Wi-Fi hotspot access. Therefore, Wi-Fi positioning technology is one of the most widely used indoor positioning algorithms, and navigation errors obtained using Wi-Fi positioning do not accumulate over time. Wi-Fi positioning results can be combined with inertial navigation to correct errors in the latter.

[0004] Optimal filter design is crucial for improving the accuracy of integrated navigation, and the Kalman filter (KF) has been widely used. In traditional integrated algorithms, the Kalman filter is typically constructed indirectly. It utilizes the inertial navigation mechanism, estimating the error state and sensor system errors through error measurement based on auxiliary sensors. For example, Zhang Yushuai et al. (2022) used an extended Kalman filter zero-velocity correction algorithm to correct navigation estimation errors in their paper "A Pedestrian Indoor Positioning Method Based on Cascaded Filtering"; Bai Nan et al. (2020) used adaptive error compensation based on motion velocity and up / down tracking based on step detection in their paper "A High-Precision and Low-Cost IMU-Based Indoor Pedestrian Positioning Technique" to improve positioning accuracy. In this strategy, the inertial sensor uses a priori error models. This strong dependence on prior inertial error models inevitably limits the use of low-cost inertial sensors (typically MEMS-IMUs), because the time-varying noise model of low-cost IMUs has high temperature sensitivity and dynamic excitation, which is extremely detrimental to the integrated system. In traditional combined algorithms, whether using indirect or direct methods, the system model in Kalman filtering is based on a priori inertial error models. In other words, the Kalman filtering process between two adjacent measurement cycles does not perform measurement updates; the IMU measurements are only used for free inertial navigation calculations. Therefore, the two methods are essentially the same. However, in indoor navigation applications, it is inevitable that Wi-Fi signals will be interrupted or unable to provide location information. Severe drift in MEMS-IMU system errors can easily lead to unacceptable free inertial navigation solutions between two auxiliary measurement updates. Summary of the Invention

[0005] The problem addressed by this invention is how to avoid the severe drift of system errors of MEMS-IMU on the free inertial navigation solution when using MEMS-IMU for indoor positioning.

[0006] To address the above problems, this invention provides an indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi, comprising:

[0007] Step 1: Pedestrians walk while holding a mobile device, and the acceleration amplitude of the accelerometer, the angular velocity of the gyroscope, and the three-axis data of the magnetometer are collected from the IMU of the mobile device.

[0008] Step 2: Establish an indoor pedestrian positioning system model based on the acceleration amplitude of the accelerometer, the angular velocity of the gyroscope, and the three-axis data of the magnetometer. This system model includes a position recursive model for indoor pedestrian positioning based on the pedestrian dead reckoning algorithm, a pedestrian attitude recursive model based on the Euler angles of the carrier coordinate system relative to the navigation coordinate system, a pedestrian angular velocity recursive model based on the Gauss-Markov process, and a system error model for the gyroscope in the IMU.

[0009] Step 3: Establish a measurement model for indoor pedestrian positioning. The measurement model includes using the output angular rate of the gyroscope in the IMU as a continuous-time observation to establish a gyroscope-based measurement model; and using the distance between the Wi-Fi access point and the mobile device collected at a fixed frequency as an auxiliary observation to establish a Wi-Fi-based measurement model.

[0010] Step 4: Based on the established system model and measurement model, establish a linearized system model and a linearized measurement model described by the extended Kalman filter, respectively;

[0011] Step 5: Estimate the state data of indoor pedestrian positioning based on the extended Kalman filter algorithm.

[0012] The beneficial effects of this invention are as follows: The position recursion model established using the pedestrian dead reckoning algorithm fundamentally alleviates the impact of severe drift of low-cost MEMS-IMU system errors on the free inertial navigation solution. Simultaneously, by establishing a gyroscope-based measurement model and a Wi-Fi-based measurement model, the observed values ​​directly and independently participate in the measurement update of the extended Kalman filter algorithm, thereby avoiding the severe dependence of the inertial navigation mechanism on the IMU output in traditional integrated strategies. Furthermore, based on the selection of IMU measurement values, the system error model established for the gyroscope in the IMU can significantly reduce the impact of the original IMU output noise, as both the original data and the system error will participate in the update of the extended Kalman filter.

[0013] Preferably, step 2, which involves establishing a recursive location model for indoor pedestrian positioning based on a pedestrian dead reckoning algorithm, specifically includes:

[0014] Step 201A: Determine the reference coordinate system. Use the carrier coordinate system of the pedestrian's initial position as the reference coordinate system for the position coordinates.

[0015] Step 202A: Obtain the vertical accelerometer output value 'a' by analyzing the peak and trough values ​​in the acceleration amplitude. vert. And based on the output value a of the vertical accelerometer vert. The step size information is calculated using the following formula:

[0016]

[0017] In the formula, These are the peak and trough values ​​of the vertical accelerometer output, respectively, and κ is the step size estimation coefficient obtained through discrete training.

[0018] Step 203A: Calculate the step size information L k By projecting the heading angle onto the navigation coordinate system and employing a pedestrian dead reckoning algorithm, the position increment of the pedestrian in the navigation coordinate system from time k to time k+1 is calculated:

[0019]

[0020] In the formula, ΔE k+1 and ΔN k+1 ψ represents the pedestrian position increments along the east and north directions at time k+1, respectively; k Let be the heading angle at time k;

[0021] Step 204A: Calculate the transformation matrix C between the pedestrian and the reference coordinate system at time k. t :

[0022]

[0023] In the formula, ψ0 is the initial heading angle; ψ k It is the heading angle at time k;

[0024] Step 205A: Transform the position increment calculated in step 203A from the navigation coordinate system to the vehicle coordinate system, and then perform the transformation using matrix C. t Transforming to the reference coordinate system, we obtain the recursive model for calculating the pedestrian's position relative to the reference coordinate system at time k+1:

[0025]

[0026] In the formula, r = (xyz) T The position vector represented using three-axis coordinates in the reference coordinate system; w r Gaussian white noise for the position vector; is the direction cosine matrix from the navigation coordinate system to the vehicle coordinate system, where n is the abbreviation for the navigation coordinate system and b is the abbreviation for the vehicle coordinate system.

[0027] Preferably, the pedestrian posture recursive model established in step 2 based on the Euler angles of the carrier coordinate system relative to the navigation coordinate system is as follows:

[0028]

[0029] In the formula, C 3×3 This is the coefficient matrix of the attitude recursion model: Where p, γ, and ψ are the pedestrian's attitude angles, which include pitch angle, roll angle, and yaw angle; The angular velocity in the carrier coordinate system. for Components along the three axes; angular acceleration This represents the process noise in the attitude recursion model.

[0030] Preferably, the pedestrian angular velocity recursive model established in step 2 based on the Gauss-Markov process is as follows:

[0031]

[0032] In the formula, Δt is the time interval; T x ,T y ,T z These are the time correlation coefficients of the first-order Markov model; It is the independent white noise of the angular velocity recursion model; It is angular velocity The non-zero mean;

[0033] Preferably, the system error model established for the gyroscope in the IMU in step 2 is as follows:

[0034]

[0035]

[0036] In the formula, b g and s g These are the drift error and scale factor error of the gyroscope, respectively. This is the white noise vector for the system error model.

[0037] Preferably, in step 3, the output angular rate of the gyroscope in the IMU is used as an observation in continuous time, and a measurement model based on the gyroscope is established as follows:

[0038]

[0039] In the formula, b g S represents the drift error of the gyroscope. g The scaling factor error matrix, (s gx ,s gy ,s gz ) represents the scaling factor error s g Components on the three axes; Δ g Gaussian white noise is used for the gyroscope measurement vector.

[0040] Preferably, in step 3, the distance between the Wi-Fi access point (AP) and the mobile device is collected at a fixed frequency as an auxiliary observation, and a Wi-Fi-based measurement model is established as follows:

[0041]

[0042] In the formula, (x n ,y n ,z n ) represents the position of the nth AP point, d i (i = 1, 2, ..., n) represents the distance between the i-th AP point and the mobile device; (x, y, z) represents the coordinates of the mobile device's position recursive model r on the three axes.

[0043] Preferably, in step 4, the linearized system model described by the extended Kalman filter is established as follows:

[0044] X k =Φ k X k-1 +Γ k W k

[0045] In the formula, X k Φ is the state vector of the system model; k W is the state transition matrix; k Γ is the process noise vector of the system model; k For W k The coefficient matrix;

[0046] In step 4, the linearized measurement model based on the extended Kalman filter, established according to the measurement model, is as follows:

[0047] Z k =H k X k +Δ k

[0048] In the formula, Z k H represents the measurement vector of the measurement model. k For measurement matrix; Δ k This is the measurement noise vector of the measurement model.

[0049] Preferably, step 5, which involves estimating the indoor pedestrian positioning state data using the extended Kalman filter algorithm, specifically includes:

[0050] Step 501: One-step prediction of the system model state vector and its variance matrix:

[0051]

[0052] Step 502: Optimal estimation of the system model state vector and its variance matrix:

[0053]

[0054] Step 503, System Innovation and its Variance Matrix:

[0055]

[0056] Step 504, Gain Matrix:

[0057] Attached Figure Description

[0058] Figure 1 This is the pedestrian navigation coordinate system used in this invention;

[0059] Figure 2 This is a schematic diagram of the pedestrian position increment calculation principle used in this invention;

[0060] Figure 3 This is a flowchart illustrating the present invention;

[0061] Figure 4 This is an example of inertial navigation data acquisition in a specific embodiment of the present invention;

[0062] Figure 5 This refers to Wi-Fi access point information in a specific embodiment of the present invention;

[0063] Figure 6 These are the actual location coordinates of the pedestrian in the specific embodiment of this invention, which are used to design the walking trajectory.

[0064] Figure 7 This is an acceleration curve of a pedestrian's movement process in a specific embodiment of the present invention;

[0065] Figure 8 This refers to the attitude estimation during MEMS-IMU standalone localization in Experiment 1 of this invention.

[0066] Figure 9 This is the position estimation when MEMS-IMU is used for independent positioning in Experiment 1 of this invention;

[0067] Figure 10 This refers to the position error when the MEMS-IMU is used for positioning alone in Experiment 1 of this invention;

[0068] Figure 11 This refers to the attitude estimation during MEMS-IMU and Wi-Fi combined localization in Experiment 2 of this invention.

[0069] Figure 12This refers to the position estimation during MEMS-IMU and Wi-Fi combined positioning in Experiment 2 of this invention;

[0070] Figure 13 This refers to the position error during MEMS-IMU and Wi-Fi combined positioning in Experiment 2 of this invention;

[0071] Figure 14 This is a comparison of attitude estimation for the two positioning methods in Experiment 1 and Experiment 2 of this invention;

[0072] Figure 15 This is a comparison of the positional errors of the two positioning methods in Experiment 1 and Experiment 2 of this invention. Detailed Implementation

[0073] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0074] An indoor pedestrian localization method combining MEMS-IMU and Wi-Fi, such as Figure 3 As shown, it includes:

[0075] Step 1: The pedestrian walks with a mobile device in hand, and the acceleration amplitude of the accelerometer, the initial attitude angle of the gyroscope, and the three-axis data of the magnetometer are collected from the IMU of the mobile device.

[0076] Step 2: Based on the collected accelerometer acceleration amplitude, gyroscope angular velocity, and magnetometer triaxial data, establish a system model for indoor pedestrian positioning. This system model includes a position recursive model for indoor pedestrian positioning based on a pedestrian dead reckoning algorithm, a pedestrian attitude recursive model based on Euler angles of the vehicle coordinate system relative to the navigation coordinate system, a pedestrian angular velocity recursive model based on a Gauss-Markov process, and a system error model for the gyroscope in the IMU; wherein:

[0077] The specific steps of establishing a recursive location model for indoor pedestrian positioning based on pedestrian dead reckoning algorithms include:

[0078] Step 201A: Determine the reference coordinate system; such as Figure 1 As shown, in this embodiment, the body coordinate frame (B-frame) of the pedestrian's initial position is used as the reference coordinate system for the position coordinates;

[0079] Step 202A, as follows Figure 7 As shown, for each step a pedestrian takes, the acceleration amplitude output by the accelerometer includes a peak value and a trough value. In this embodiment, the vertical accelerometer output value 'a' is obtained by using the peak and trough values ​​in the acceleration amplitude. vert.To detect changes in stride length, and based on the output value a of the vertical accelerometer. vert. The step size information is calculated using the following formula:

[0080]

[0081] In the formula, These represent the peak and trough values ​​of the vertical accelerometer output, respectively; κ is the step size estimation coefficient obtained through discrete training, and in this embodiment, κ = 0.5;

[0082] Step 203A: Calculate the step size information L k By projecting the heading angle onto the navigation coordinate system, a pedestrian dead reckoning algorithm is used. The calculation principle is as follows: Figure 2 As shown, calculate the pedestrian's position increment in the navigation coordinate system from time k to time k+1:

[0083]

[0084] In the formula, ΔE k+1 and ΔN k+1 ψ represents the pedestrian position increments along the east and north directions at time k+1, respectively; k L is the heading angle at time k; k This refers to the step size information from time k to time k+1;

[0085] Step 204A: Calculate the transformation matrix C between the pedestrian and the reference coordinate system at time k. t :

[0086]

[0087] In the formula, ψ0 is the initial heading angle; ψ k It is the heading angle at time k; in order to verify the subsequent experimental analysis, this embodiment designs a walking trajectory of "straight line + right-angle right turn", and the transformation matrix C of the "straight line" motion segment is... t The transformation matrix of the "right-angle right turn" motion segment is an identity matrix.

[0088] Step 205A: Transform the position increment calculated in Step 203A from the navigation coordinate frame (n-frame) to the vehicle coordinate frame, and then perform a transformation using matrix C. t Transforming to the reference coordinate system, we obtain the recursive model for calculating the pedestrian's position relative to the reference coordinate system at time k+1:

[0089]

[0090] In the formula, r = (xyz)T The position vector represented using three-axis coordinates in the reference coordinate system; w r Gaussian white noise for the position vector; Here, is the direction cosine matrix from the navigation coordinate system to the vehicle coordinate system, where n is the abbreviation for the navigation coordinate system and b is the abbreviation for the vehicle coordinate system. In this embodiment, when calculating the position recursion model, it is necessary to use the 2×1 dimensional position increment. Extend to 3×1 dimension, that is, set the corresponding third dimension position quantity to 0, and only take the two-dimensional coordinates (x,y) when calculating the position coordinates;

[0091] The pedestrian's posture recursive model is established based on the Euler angles of the vehicle coordinate system relative to the navigation coordinate system as follows:

[0092]

[0093] In the formula, C 3×3 This is the coefficient matrix of the attitude recursion model: Where p, γ, and ψ are the pedestrian's attitude angles, and the initial values ​​of p, γ, and ψ are directly determined by the three-axis data output of the magnetometer in the mobile device at the initial moment. The attitude angles include pitch angle, roll angle, and yaw angle. The angular velocity in the carrier coordinate system. for Components along the three axes; angular acceleration This refers to the process noise in the attitude recursion model.

[0094] The recursive model for pedestrian angular velocity based on the Gauss-Markov process is as follows:

[0095]

[0096] In the formula, Δt is the time interval; T x ,T y ,T z These are the time correlation coefficients of the first-order Markov model; It is the independent white noise of the angular velocity recursion model; It is angular velocity The non-zero mean is used in this embodiment to calculate the angular velocity at the previous moment. replace;

[0097] The systematic error model established for the gyroscope in the IMU is as follows:

[0098]

[0099]

[0100] In the formula, b g and s gThese are the drift error and scale factor error of the gyroscope, respectively. The system error model uses a white noise vector.

[0101] Step 3: Establish a measurement model for indoor pedestrian positioning. This measurement model includes establishing a gyroscope-based measurement model by using the output angular rate of the gyroscope in the IMU as a continuous-time observation; and establishing a Wi-Fi-based measurement model by using the distance between the Wi-Fi access point (AP) and the mobile device as an auxiliary observation, wherein:

[0102] The output angular rate of the gyroscope in the IMU is used as a continuous-time observation. The measurement model based on the gyroscope is established as follows:

[0103]

[0104] In the formula, b g S represents the drift error of the gyroscope. g The scaling factor error matrix, (s gx ,s gy ,s gz ) represents the scaling factor error s g Components on the three axes; Δ g Gaussian white noise for gyroscope measurement vectors;

[0105] Using fixed-frequency data acquisition of the distance between the Wi-Fi access point (AP) and the mobile device as an auxiliary observation, a Wi-Fi-based measurement model is established as follows:

[0106]

[0107] In the formula, (x n ,y n ,z n ) represents the location of the nth access point (AP). To facilitate determining the actual location of the AP, this embodiment uses a smartphone hotspot as the Wi-Fi signal, and sets n = 5, d i (i = 1, 2, ..., n) represents the distance between the i-th access point (AP) and the mobile device, which is obtained in this embodiment based on the strength of the Wi-Fi signal. In the formula, RSS represents the strength of the Wi-Fi signal received by the smartphone based on its own Wi-Fi receiver module, such as... Figure 5 As shown, A and η are both propagation parameters. In this embodiment, the access frequency of the Wi-Fi signal is f = 0.625; the propagation parameters A = 33 and η = 2; (x, y, z) represent the coordinates of the mobile device's position recursive model r on the three axes.

[0108] Step 4: Based on the established system model and measurement model, establish a linearized system model and a linearized measurement model described by the extended Kalman filter, respectively; where:

[0109] Based on the system model, the linearized system model described by the extended Kalman filter is established as follows:

[0110] X k =Φ k X k-1 +Γ k W k

[0111] In the formula, X k In this embodiment, the state vector of the system model is... Φ k W is the state transition matrix; k The process noise vector of the system model is shown in this embodiment. And W k ~N(0,Q) k );Γ k For W k The coefficient matrix;

[0112] Based on the measurement model, a linearized measurement model described by the extended Kalman filter is established as follows:

[0113] Z k =H k X k +Δ k

[0114] In the formula, Z k In this embodiment, the measurement vector of the measurement model is... H k For measurement matrix; Δ k Let be the measurement noise vector of the measurement model, and Δ k ~N(0,R k ); Here, W k and Δ k They are unrelated;

[0115] Step 5: Estimate the state data for indoor pedestrian positioning using the extended Kalman filter algorithm, specifically including:

[0116] Step 501: One-step prediction of the system model state vector and its variance matrix:

[0117]

[0118] Step 502: Optimal estimation of the system model state vector and its variance matrix:

[0119]

[0120] Step 503, System Innovation and its Variance Matrix:

[0121]

[0122] Step 504, Gain Matrix:

[0123]

[0124] Experimental Analysis:

[0125] The following experimental analysis is conducted on the positioning performance and effectiveness of an indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi.

[0126] Actual data collection:

[0127] Pedestrians use a mobile device to walk along a pre-defined path, obtaining the position coordinates of the corresponding path. In this embodiment, the mobile device is an Android smartphone. The walking path designed in this embodiment uses a "straight line + right-angle right turn" path. The true value of the position coordinates of the corresponding path is obtained, such as... Figure 6 As shown; and by using an inertial navigation data collector, three-axis data from the accelerometer, gyroscope, and magnetometer in the smartphone were collected, such as... Figure 4 As shown.

[0128] Experiment 1:

[0129] This experiment only uses data from the built-in MEMS-IMU of a smartphone to calculate the positioning result of the "straight line + right-angle right turn" trajectory. The positioning result includes, for example: Figure 8 The attitude angle estimation shown is as follows: Figure 9 The position coordinate estimate shown is obtained by subtracting the position coordinate estimate from the true position coordinate value. Figure 10 The position error curve shown illustrates the measurement update process where Wi-Fi signal data is not used for assisted positioning, and only the angular rate output by the gyroscope is used as the observed value for filtering. Figure 8-10 It can be seen that the positioning error using the smartphone's built-in IMU increases slowly over time, reaching as low as 0.1m during a 16-second walking process. Figure 6 The true position coordinates (total walking distance approximately 10m) are within the allowable range, therefore this scheme is an effective positioning scheme.

[0130] Experiment 2:

[0131] This experiment uses Wi-Fi signal data and real-time data from a smartphone's built-in MEMS-IMU to calculate the positioning result of a "straight line + right-angle right turn" trajectory. The positioning result includes, for example: Figure 11 The attitude angle estimation shown and as Figure 12 The position coordinate estimate shown is obtained by subtracting the position coordinate estimate from the true position coordinate value. Figure 13 The position error curve shown; according to Figure 5 The MAC address is used to determine the device name. The distance d between each AP point and the pedestrian carrying the mobile device is inferred from the signal strength. When all three selected mobile devices have valid location information, a set of distances (d1, d2, d3) is selected as auxiliary observation. The access frequency of the observation data is determined according to the number of data sets collected at different times. In this experiment, 10 sets of Wi-Fi distance data were collected at different times during the 16-second walking process. The data were then filtered and updated in the pedestrian positioning algorithm at a frequency of 0.625.

[0132] Comparing Experiment 1 and Experiment 2: From Figure 14 It can be seen that with the assistance of Wi-Fi information, the system's attitude estimation is much more stable, which is consistent with the attitude changes of pedestrians under the designed path (pitch and roll fluctuate less around zero, and heading angle also fluctuates less near the initial value). According to the position recursion formula, the attitude matrix is ​​the key to position coordinate transformation, and the accuracy of attitude angle estimation directly affects the accuracy of position estimation.

[0133] Figure 15 The results show that the positioning error using the built-in IMU increases slowly over time, eventually exceeding 0.1m. However, when MEMS-IMU and Wi-Fi are used for positioning, the positioning error is still less than 0.1m. This demonstrates that Wi-Fi information can improve the accuracy of pedestrian positioning and assist MEMS inertial sensors in completing indoor pedestrian navigation and positioning tasks more accurately.

[0134] While the disclosure is as stated above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of this disclosure, and all such changes and modifications will fall within the protection scope of this invention.

Claims

1. An indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi, characterized in that, include: Step 1: Pedestrians walk while holding a mobile device, and the acceleration amplitude of the accelerometer, the angular velocity of the gyroscope, and the three-axis data of the magnetometer are collected from the IMU of the mobile device. Step 2: Establish an indoor pedestrian positioning system model based on the acceleration amplitude of the accelerometer, the angular velocity of the gyroscope, and the three-axis data of the magnetometer. The system model includes a position recursive model for indoor pedestrian positioning based on a pedestrian dead reckoning algorithm, a pedestrian attitude recursive model based on the Euler angles of the carrier coordinate system relative to the navigation coordinate system, a pedestrian angular velocity recursive model based on a Gauss-Markov process, and a system error model for the gyroscope in the IMU. Step 3: Establish a measurement model for indoor pedestrian positioning. The measurement model includes using the output angular rate of the gyroscope in the IMU as a continuous-time observation to establish a gyroscope-based measurement model; and using the distance between the Wi-Fi access point and the mobile device collected at a fixed frequency as an auxiliary observation to establish a Wi-Fi-based measurement model. Step 4: Based on the established system model and measurement model, establish a linearized system model and a linearized measurement model described by the extended Kalman filter, respectively; Step 5: Estimate the state data of indoor pedestrian positioning based on the extended Kalman filter algorithm.

2. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, Step 2, which involves establishing a recursive location model for indoor pedestrian positioning based on a pedestrian dead reckoning algorithm, specifically includes: Step 201A: Determine the reference coordinate system. Use the carrier coordinate system of the pedestrian's initial position as the reference coordinate system for the position coordinates. Step 202A: Obtain the vertical accelerometer output value 'a' by analyzing the peak and trough values ​​in the acceleration amplitude. vert. And based on the output value a of the vertical accelerometer vert. The step size information is calculated using the following formula: In the formula, These are the peak and trough values ​​of the vertical accelerometer output, respectively, and κ is the step size estimation coefficient obtained through discrete training. Step 203A: Calculate the step size information L k By projecting the heading angle onto the navigation coordinate system and employing a pedestrian dead reckoning algorithm, the position increment of the pedestrian in the navigation coordinate system from time k to time k+1 is calculated: In the formula, ΔE k+1 and ΔN k+1 ψ represents the pedestrian position increments along the east and north directions at time k+1, respectively; k Let be the heading angle at time k; Step 204A: Calculate the transformation matrix C between the pedestrian and the reference coordinate system at time k. t : In the formula, ψ0 is the initial heading angle; ψ k It is the heading angle at time k; Step 205A: Transform the position increment calculated in step 203A from the navigation coordinate system to the vehicle coordinate system, and then perform the transformation using matrix C. t Transforming to the reference coordinate system, we obtain the recursive model for calculating the pedestrian's position relative to the reference coordinate system at time k+1: In the formula, r = (xyz) T The position vector represented using three-axis coordinates in the reference coordinate system; w r Gaussian white noise for the position vector; Let n be the direction cosine matrix from the navigation coordinate system to the vehicle coordinate system, where n is the abbreviation for the navigation coordinate system and b is the abbreviation for the vehicle coordinate system.

3. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, In step 2, the pedestrian's posture recursive model is established based on the Euler angles of the vehicle coordinate system relative to the navigation coordinate system as follows: In the formula, C 3×3 This is the coefficient matrix of the attitude recursion model: Where p, γ, and ψ are the pedestrian's attitude angles, which include pitch angle, roll angle, and yaw angle; The angular velocity in the carrier coordinate system. for Components along the three axes; angular acceleration This represents the process noise in the attitude recursion model.

4. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, The pedestrian angular velocity recursive model established in step 2 based on the Gauss-Markov process is as follows: In the formula, Δt is the time interval; T x ,T y ,T z These are the time correlation coefficients of the first-order Markov model; It is the independent white noise of the angular velocity recursion model; It is angular velocity The non-zero mean.

5. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, The system error model established for the gyroscope in the IMU in step 2 is as follows: In the formula, b g and s g These are the drift error and scale factor error of the gyroscope, respectively. This is the white noise vector for the system error model.

6. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, In step 3, the output angular rate of the gyroscope in the IMU is used as an observation in continuous time, and a measurement model based on the gyroscope is established as follows: In the formula, b g S represents the drift error of the gyroscope. g The scaling factor error matrix, (s gx ,s gy ,s gz ) represents the scaling factor error s g The components on the three axes; Δg is the Gaussian white noise of the gyroscope measurement vector.

7. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, In step 3, the distance between the Wi-Fi access point (AP) and the mobile device is collected at a fixed frequency as an auxiliary observation. The Wi-Fi-based measurement model is established as follows: In the formula, (x n ,y n ,z n ) represents the position of the nth AP point, d i (i = 1, 2, ..., n) represents the distance between the i-th AP point and the mobile device; (x,y,z) represents the coordinates of the mobile device's position recursive model r on the three axes.

8. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 1, characterized in that, In step 4, the linearized system model described by the extended Kalman filter is established based on the system model as follows: X k =Φ k X k-1 +C k W k In the formula, X k Φ is the state vector of the system model; k W is the state transition matrix; k Γ is the process noise vector of the system model; k For W k The coefficient matrix; In step 4, the linearized measurement model based on the extended Kalman filter, established according to the measurement model, is as follows: Z k =H k X k +Δ k In the formula, Z k H represents the measurement vector of the measurement model. k For measurement matrix; Δ k This is the measurement noise vector of the measurement model.

9. The indoor pedestrian positioning method combining MEMS-IMU and Wi-Fi according to claim 8, characterized in that, Step 5, which estimates the state data for indoor pedestrian positioning using the extended Kalman filter algorithm, specifically includes: Step 501: One-step prediction of the system model state vector and its variance matrix: Step 502: Optimal estimation of the system model state vector and its variance matrix: Step 503, System Innovation and its Variance Matrix: Step 504, Gain Matrix:

Images