Multi-AP cooperative topological positioning method and system based on particle filtering

Abstract

The invention discloses a multi-AP cooperative topology positioning method and system based on particle filtering, and the method comprises the steps: obtaining the motion information of user equipment (UE) through an inertial measurement unit (IMU) on the basis of channel state information (CSI) provided by a plurality of wireless access points (AP) with known positions; s2, a correction UE track filter combines the UE motion information and CSI measurement data in the step S1 to update the weight and the position of a particle in real time, and corrects the position estimation of the UE; the estimated AP position filter takes the estimated position of the UE as input and updates the position estimation of the unknown AP in combination with the CSI between the unknown AP and the UE; and finally restoring the AP network topology structure through weighted average of the positions of all particles. According to the invention, the problem of position error accumulation caused by motion signal noise in the existing IMU positioning system can be effectively solved, so that the positioning precision and robustness are improved, and the method is suitable for various application scenes such as indoor positioning, automatic driving, robot navigation and the like.

Description

Technical Field

[0001] The present invention belongs to the field of communications, relates to topology positioning technology, and particularly relates to a multi-AP collaborative topology positioning method and system based on particle filtering. Background Art

[0002] With the rapid development of technologies such as smart devices, the Internet of Things, autonomous driving, and robot navigation, precise position tracking and positioning technologies have been widely applied in many application scenarios. As an important positioning method, indoor positioning has become one of the research hotspots, especially in indoor environments where Global Positioning System (GPS) signals cannot cover, and the demand for positioning technology is particularly urgent. Traditional indoor positioning technologies mainly include positioning methods based on wireless communication technologies such as Wi-Fi, Bluetooth, ultra-wideband, and infrared signals, as well as sensor-based positioning methods such as inertial navigation systems and visual inertial odometers. However, the existing indoor positioning technologies still have the following main problems:

[0003] Problem 1: Sensor noise problem: Most sensor-based positioning methods, especially IMU positioning methods, usually rely on sensors such as accelerometers and gyroscopes to estimate the motion trajectory of an object. However, the noise and drift of these sensors will cause the error of position estimation to gradually accumulate over time. For example, when the acceleration signal obtains position information through two integrations, any small error will cause a large cumulative error, resulting in position deviation and even possible failure.

[0004] Problem 2: Multipath effect and signal interference: Radio signal-based positioning methods, such as Wi-Fi and Bluetooth, are easily affected by the multipath effect and signal interference. The multipath effect is caused by the reflection, refraction, and scattering of signals during propagation, resulting in a deviation in the time when the received signal reaches the receiver, thereby affecting the positioning accuracy. The signal interference problem is mainly manifested as the existence of multiple different types of signal sources in the environment, interfering with each other, which reduces the robustness of the positioning system.

[0005] Problem 3: Accuracy and stability problems of traditional positioning methods: Traditional positioning methods based on ranging or Angle of Arrival (AOA) information, such as estimating the position by calculating the distance or angle from an object to wireless access points such as AP1, AP2, and AP3, can achieve a certain accuracy of positioning under ideal conditions. However, in complex environments, such as high-rise buildings, dense obstacles, signal interference, etc., the positioning accuracy and stability are often affected. Especially for the position of unknown APs, it is difficult to achieve a high estimation accuracy simply relying on traditional methods.

[0006] Problem 4: Insufficient joint optimization of IMU and AP angle information: Although the motion signals obtained from the IMU and the CSI obtained from the AP each have their own advantages in positioning, traditional methods often process the motion information and CSI independently, failing to fully combine the advantages of both, resulting in insufficient positioning accuracy. Summary of the Invention

[0007] Object of the Invention: In order to overcome the deficiencies in the prior art, a multi-AP collaborative topology positioning method and system based on particle filtering are provided. By combining the motion information provided by the IMU and the CSI provided by the AP, and utilizing the synergistic effect of two particle filters, the positions of the UE and the unknown-position AP are accurately estimated. The method of the present invention can effectively solve the problem of position error accumulation caused by motion signal noise in the existing IMU positioning system, thereby improving the positioning accuracy and robustness, and is applicable to various application scenarios such as indoor positioning, autonomous driving, and robot navigation.

[0008] Technical Solution: To achieve the above object, the present invention provides a multi-AP collaborative topology positioning method based on particle filtering, including the following steps:

[0009] S1: Based on the channel state information CSI provided by multiple wireless access points AP at known positions, obtain the motion information of the user equipment UE through the inertial measurement unit IMU;

[0010] S2: The corrected UE trajectory filter combines the UE motion information and CSI measurement data in step S1, updates the weights and positions of the particles in real time, and corrects the position estimate of the UE;

[0011] S3: The estimated AP position filter takes the estimated position of the UE as input, combines the CSI between the unknown AP and the UE, and updates the position estimate of the unknown AP;

[0012] S4: By weighted averaging the positions of all particles, finally restore the AP network topology structure.

[0013] Further, the operation of the corrected UE trajectory filter in step S2 includes:

[0014] A1: Particle initialization: Randomly initialize multiple particles near the initial position of the UE. Each particle represents a possible motion state of the UE, and randomly initialize the motion of the particles;

[0015] A2: Particle propagation: According to the dynamic update rules of motion information such as acceleration and velocity, update the motion information of each particle to simulate the motion of the UE;

[0016] A3: Measurement update: According to the CSI measurement values provided by the known AP, calculate the predicted CSI of each particle, compare it with the actual measurement value, and update the weights of the particles;

[0017] A4: Normalized Weight: Normalize the weights of the particles to ensure that the sum of the weights of all particles is 1;

[0018] A5: Resampling: If the weights of most particles are close to zero, the diversity of the particles becomes very low, which will lead to inaccurate estimation results. Therefore, resampling is performed to ensure uniform distribution of the particles and avoid particle degradation;

[0019] A6: Position Estimation: Obtain the estimated motion state of the UE by weighted averaging the positions of all particles.

[0020] Furthermore, in step A2, in order to simulate the true motion trajectory of the UE, a uniformly accelerated motion model is adopted in the X direction, and a uniform linear motion is adopted in the Y direction. The position and velocity of the UE are updated through the following formulas:

[0021] user_pos_true(t + 1)

[0022] = user_pos_true(t) + user_vel_true(t) × T + 0.5 × acceleration_true × T 2 At the same time, simulate the motion data read by the IMU, add noise interference, and the update formula is as follows:

[0023] user_pos_false(t + 1) = user_pos_false(t) + user_vel_false(t) × T + 0.5 × a meas × T 2

[0024] where a meas is the measured acceleration value after adding noise.

[0025] Furthermore, in step A2, at each time step, the particle filter updates the velocity and position of the particles through the motion model; the motion model is as follows:

[0026] px_user(3, :) = px_user(3, :) + a particles × T

[0027] px_user(1, :) = px_user(1, :) + px_user(3, :) × T + 0.5 × a particles × T 2 .

[0028] Furthermore, the calculation formula for predicting CSI (taking AOA as an example) in step A3 is:

[0029] theta_pred_AP1 = atan2(AP1_pos(2) - px_user(2, :), AP1_pos(1) - px_user(1, :));

[0030] theta_pred_AP2 = atan2(AP2_pos(2) - px_user(2, :), AP2_pos(1) - px_user(1, :));

[0031] Update the weights of the particles: The likelihood function from the Gaussian distribution represents the joint likelihood of the independent observation errors of AP1 and AP2: pq_user = exp(-0.5 * (vy_AP1 / v_std).^2).* exp(-0.5 * (vy_AP2 / v_std).^2);

[0032] where vy_AP1 and vy_AP2 are the differences between the true CSI of the user relative to the known AP and the CSI of the particle relative to the known AP; v_std is the standard deviation of the observation error.

[0033] Furthermore, the operation of estimating the AP position filter in step S3 includes:

[0034] B1: Particle initialization: The AP position filter estimates the possible position range of the unknown AP by randomly initializing particles, and the particle positions are uniformly distributed within a preset area;

[0035] B2: Particle propagation: The particles propagate according to the process noise to simulate the change of the unknown AP position; in the absence of obvious motion information, the particle positions are mainly determined by the process noise;

[0036] B3: Measurement update: According to the CSI measurement values of the unknown AP and the UE, calculate the predicted CSI of each particle and compare it with the actual measurement value to update the weights of the particles; the weight update is based on the difference between the predicted CSI and the actual CSI, and the greater the position weight of the particle, the more accurate the position estimate;

[0037] Update the weights of the particles:

[0038] pq_AP3 = exp(-0.5 * (vy_AP3_pf2 / v_std).^2);

[0039] where vy_AP3_pf2 is the difference between the true CSI of AP3 and the user and the CSI of the particle and the user.

[0040] B4: Normalize the weights: Normalize the weights of all particles to ensure that the sum of the particle weights is 1;

[0041] B5: Resampling: If most of the particle weights are close to zero, the diversity of the particles becomes very low, leading to inaccurate estimation results. Therefore, resampling is performed to ensure that the particles are concentrated in the high-weight regions and avoid particle degeneration.

[0042] B6: Location Estimation: The estimated location of the unknown AP is obtained by weighted averaging the locations of all the particles.

[0043] In the present invention, both the corrected UE trajectory filter and the estimated AP location filter use the Gaussian weighting method to weight the particles and update the particle weights using CSI.

[0044] The present invention also provides a multi-AP cooperative topology localization system based on particle filtering, including:

[0045] Multiple APs for providing CSI;

[0046] An IMU for obtaining the motion information of the UE;

[0047] A corrected UE trajectory filter for estimating the real-time location of the UE based on the motion information and CSI;

[0048] An estimated AP location filter for estimating the location information of the unknown AP based on the UE location and the CSI of the unknown AP.

[0049] According to the above content, the present invention mainly includes the following parts:

[0050] UE Location Estimation: The motion information of the UE is obtained using the IMU, and the displacement of the UE is deduced based on this information. Since there is noise in the IMU signal, a corrected UE trajectory filter is used to estimate the real location of the UE. The corrected UE trajectory filter simulates multiple particles to represent the possible locations and possible accelerations of the UE, and updates the weights of the particles by comparing the CSI provided by each particle with the known APs and the CSI provided by the actual UE with the known APs. The particle degeneration is effectively avoided through the resampling mechanism, and the UE location is corrected.

[0051] Location Estimation of Unknown AP: The UE location estimated by the corrected UE trajectory filter is used as input information and passed to the estimated AP location filter for estimating the location information of the unknown AP. The estimated AP location filter updates the weights of the particles by comparing the CSI provided by each particle with the known APs and the CSI provided by the unknown AP with the known APs, effectively avoids particle degeneration through the resampling mechanism, and finally estimates the real location of the unknown AP.

[0052] Synergistic effect of particle filters: The calibration UE trajectory filter and the estimated AP position filter work together to jointly optimize the motion information obtained from the IMU and the CSI obtained from the AP, reduce the positioning error caused by the IMU signal noise, and ensure the accurate estimation of the UE position and the unknown AP position.

[0053] Beneficial effects: Compared with the prior art, the present invention combines the motion information provided by the IMU and the CSI provided by the AP, and uses the synergistic effect of two particle filters to accurately estimate the positions of the UE and the AP at unknown positions. The method of the present invention can effectively solve the problem of position error accumulation caused by motion signal noise in the existing IMU positioning system, thereby improving the positioning accuracy and robustness, and is applicable to various application scenarios such as indoor positioning, autonomous driving, and robot navigation. Description of the drawings

[0054] Figure 1 is the flowchart of the method of the present invention;

[0055] Figure 2 is the operation flowchart of the calibration UE trajectory filter in the present invention;

[0056] Figure 3 is the operation flowchart of the estimated AP position filter in the present invention;

[0057] Figure 4 is the AOA schematic diagram in the method of the present invention;

[0058] Figure 5 is the UE trajectory error map before calibration;

[0059] Figure 6 is the UE trajectory error map after calibration;

[0060] Figure 7 is the scatter plot of the position estimation result of the unknown AP (AP3) changing with time;

[0061] Figure 8 is the graph of the position estimation error of the unknown AP (AP3) changing with time. Detailed implementation manners

[0062] The present invention will be further clarified below in conjunction with the drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, various equivalent modifications made by those skilled in the art to the present invention all fall within the scope defined by the appended claims of this application.

[0063] Embodiment 1:

[0064] This embodiment provides a multi-AP collaborative topology positioning system based on particle filtering, including:

[0065] Multiple APs, which are used to provide CSI (in this embodiment, the selected CSI is AOA);

[0066] An IMU, which is used to obtain the motion information of the UE;

[0067] A UE trajectory corrector, which is used to estimate the real-time position of the UE according to the motion information and AOA;

[0068] An estimated AP position filter, which is used to estimate the position information of an unknown AP according to the UE position and the AOA of the unknown AP.

[0069] This embodiment applies the above positioning system. In this embodiment, the known APs are AP1 and AP2, and the unknown AP is AP3. A multi-AP collaborative topology positioning method based on particle filtering is provided. As Figure 1 shown, it includes the following steps:

[0070] 1) Based on the AOA of the channel state information provided by multiple wireless access points APs at known positions, the motion information of the user equipment UE is obtained through the inertial measurement unit IMU;

[0071] System parameter initialization: including setting the sampling period T = 0.1 second, the maximum number of time steps N f = 600, and the time vector:

[0072] tim = 0:T:(N f -1)×T

[0073] At the same time, set the positions of the known APs AP1 and AP2 as follows:

[0074] AP1_pos = [0;0], AP2_pos = [20;70]

[0075] The real position of the unknown AP3 is:

[0076] AP3_pos_true = [30;30]

[0077] The initial position and velocity of the UE device are set as:

[0078] user_init_pos = [0;0], user_init_vel = [0;1]

[0079] The acceleration of the UE is set to 0.05, and acceleration noise is simulated to consider random interference in the actual environment.

[0080] 2) The UE trajectory corrector combines the UE motion information and the CSI measurement data, updates the weights and positions of the particles in real time, and corrects the UE position estimate;

[0081] Refer to Figure 2 , the operation of calibrating the UE trajectory filter includes:

[0082] A1: Particle initialization: Randomly initialize multiple particles near the initial position of the UE. Each particle represents a possible motion state of the UE, and randomly initialize the motion of the particles;

[0083] A2: Particle propagation: Update the motion information of each particle according to the dynamic update rules of motion information (such as acceleration and velocity) to simulate the motion of the UE;

[0084] A3: Measurement update: According to the AOA measurement values provided by the known APs, calculate the predicted AOA of each particle, compare it with the actual measurement value, and update the weight of the particle;

[0085] The calculation formula for the predicted AOA is:

[0086] theta_pred_AP1 = atan2(AP1_pos(2)-px_user(2,:), AP1_pos(1)-px_user(1,:));

[0087] theta_pred_AP2 = atan2(AP2_pos(2)-px_user(2,:), AP2_pos(1)-px_user(1,:));

[0088] Update the weight of the particle: The likelihood function from the Gaussian distribution represents the joint likelihood of the independent observation errors of AP1 and AP2: pq_user = exp(-0.5*(vy_AP1 / v_std).^2).*exp(-0.5*(vy_AP2 / v_std).^2);

[0089] where vy_AP1 and vy_AP2 are the differences between the true AOA of the user relative to the known AP and the AOA of the particle relative to the known AP; v_std is the standard deviation of the observation error.

[0090] A4: Normalize weights: Normalize the weights of the particles to ensure that the sum of the weights of all particles is 1;

[0091] A5: Resampling: If the weights of most particles are close to zero, the diversity of the particles becomes very low, which will lead to inaccurate estimation results. Therefore, resampling is performed to ensure uniform distribution of the particles and avoid particle degradation;

[0092] A6: Position estimation: Obtain the estimated motion state of the UE by weighted averaging the positions of all particles.

[0093] The calibration UE trajectory filter uses motion information such as velocity, acceleration, and AOA measurement data through the above steps to continuously correct the position estimation of particles, thereby achieving high-precision UE positioning.

[0094] In this embodiment, in order to simulate the real motion trajectory of the UE, a uniformly accelerated motion model is adopted in the X direction, a uniform linear motion is adopted in the Y direction, and the position and velocity of the UE are updated through the following formula:

[0095] user_pos_true(t + 1)

[0096] = user_pos_true(t) + user_vel_true(t) × T + 0.5 × acceleration_true × T 2 At the same time, simulate the motion data read by the IMU, add noise interference, and the update formula is as follows:

[0097] user_pos_false(t + 1) = user_pos_false(t) + user_vel_false(t) × T + 0.5 × a meas × T 2

[0098] where a meas is the acceleration measurement value after adding noise;

[0099] The main function of the calibration UE trajectory filter is to combine the acceleration information a meas provided by the IMU and the AOA information of the known AP to perform real-time estimation and correction of the UE position. The specific process is as follows: First, initialize the particles. Each particle represents the possible acceleration and possible position of the UE. Set the number of particles N p = 3000, and randomly initialize the particles near the initial position of the UE. Each particle contains information such as position, velocity, and acceleration. Second, update the particle state. At each time step, the particle filter updates the velocity and position of the particles through the motion model. The motion model is as follows:

[0100] px_user(3, :) = px_user(3, :) + a particles × T

[0101] px_user(1, :) = px_user(1, :) + px_user(3, :) × T + 0.5 × a particles × T 2

[0102] Then perform particle weighting and position estimation, referring to Figure 4, using the true AOA information provided by AP1 and AP2, calculate the observation error, update the particle weights, and calculate the estimated UE position through weighted average:

[0103] xe_user = ∑(px_user(1:2, :) × pq_user)

[0104] Finally, there is particle resampling. When the number of effective particles is lower than the set threshold, the resampling process is executed to avoid particle degradation and improve the accuracy of position estimation.

[0105] S3: The estimated AP position filter takes the estimated position of the UE as input, combines the CSI between the unknown AP and the UE, and updates the estimated position of the unknown AP;

[0106] Refer to Figure 3 , the operation of the estimated AP position filter includes:

[0107] B1: Particle initialization: The estimated AP position filter randomly initializes particles within the possible position range of the unknown AP, and the particle positions are evenly distributed within the preset area;

[0108] B2: Particle propagation: The particles propagate according to the process noise, simulating the change of the unknown AP position; in the absence of obvious motion information, the particle positions are mainly determined by the process noise;

[0109] B3: Measurement update: According to the CSI measurement values between the unknown AP and the UE, calculate the predicted AOA of each particle, compare it with the actual measurement value, and update the particle weights; the weight update is based on the difference between the predicted AOA and the actual AOA. The greater the position weight of the particle, the more accurate the position estimate;

[0110] The calculation formula for the predicted AOA is:

[0111] theta_pred_AP3_pf2 = atan2(px_AP3(2, :) - xe_user(2), px_AP3(1, :) - xe_user(1));

[0112] Update the particle weights:

[0113] pq_AP3 = exp(-0.5 * (vy_AP3_pf2 / v_std).^2);

[0114] where vy_AP3_pf2 is the difference between the true AOA of AP3 and the user and the AOA of the particle and the user.

[0115] B4: Normalize weights: Normalize the weights of all particles to ensure that the sum of the particle weights is 1;

[0116] B5: Resampling: If most of the particle weights are close to zero, the diversity of the particles becomes very low, leading to inaccurate estimation results. Therefore, resampling is performed to ensure that the particles are concentrated in the high-weight regions and avoid particle degeneration.

[0117] B6: Location Estimation: The estimated location of the unknown AP is obtained by weighted averaging the locations of all particles.

[0118] The AP location estimation filter continuously updates the location estimation of the unknown AP by combining the estimated value of the UE location and the AOA measurement information between the unknown AP and the UE, ensuring that the location information of the unknown AP is accurately estimated in a dynamic environment.

[0119] In this embodiment, the main function of the AP location estimation filter is to estimate the location of AP3 by combining the true AOA information provided by AP3. The specific process is as follows: First, initialize the particles. Each particle represents a possible location of AP3, and the number of particles N p = 3000, and the particles are randomly initialized within the specified range. Secondly, particle weighting and location estimation are performed. The AP location estimation filter uses the UE location estimated by the corrected UE trajectory filter as the input. Referring to Figure 4 , the true AOA information between the UE location and AP3 and the AOA information between the UE location and the particles are used to update the particle weights, and the location estimation of AP3 is calculated by weighted averaging:

[0120] theta_true_AP3 = atan2(AP3_pos_true(2) - user_pos_true(2,nn), AP3_pos_true(1) - user_pos_true(1,nn))

[0121] ym_AP3 = theta_true_AP3 + v(3,nn), where v is the measurement noise of the AOA.

[0122] theta_pred_AP3_pf2 = atan2(px_AP3(2,:) - xe_user(2), px_AP3(1,:) - xe_user(1))

[0123] The observation error is:

[0124] vy_AP3_pf2 = ym_AP3 - theta_pred_AP3_pf2

[0125] Finally, the resampling process is performed. If the number of effective particles is insufficient, resampling is executed to ensure the stability and estimation accuracy of the particle filter.

[0126] S4: By weighted averaging the positions of all particles, the AP network topology is finally restored.

[0127] Embodiment 2:

[0128] To verify the effectiveness and effect of the present invention, in this embodiment, experiments are carried out and data analysis is performed, specifically as follows:

[0129] Figure 5 and Figure 6 shows the UE position correction effect. Before correction, the X-direction error of the UE increases significantly with time, reaching a maximum of about 5 meters. After correction, the X-direction error decreases significantly, and the maximum error is less than 0.5 meters.

[0130] Figure 7 shows the change of the AP3 position estimation result over time. As Figure 7 shown, the estimation result of the AP3 position gradually converges over time, and the color depth represents the time step. In the initial stage, the error of the position estimation is large, but as time increases, the particle filter makes the estimation result gradually tend to the true position through multiple iterations; Figure 8 is the analysis of the AP3 estimated position error. The error is large in the initial stage (about 40 meters), but under the action of the AP position estimation filter, the error quickly decreases and converges to zero, proving the effectiveness and accuracy of the method of the present invention.

[0131] It can be seen that the present invention can gradually optimize the position estimation of the UE and unknown APs in a wireless environment and finally achieve the efficient restoration of the network topology.

Claims

1. A multi-AP collaborative topology localization method based on particle filtering, characterized in that It includes the following steps: S1: Based on the channel state information CSI provided by multiple wireless access points AP at known positions, obtain the motion information of the user equipment UE through the inertial measurement unit IMU; S2: The calibrated UE trajectory filter combines the UE motion information and CSI measurement data in step S1, updates the weights and positions of the particles in real time, and corrects the position estimate of the UE; S3: The estimated AP position filter takes the estimated position of the UE as input, combines the CSI between the unknown AP and the UE, and updates the position estimate of the unknown AP; S4: By weighted averaging the positions of all particles, finally restore the AP network topology.

2. A multi-AP collaborative topology localization method based on particle filtering according to claim 1, characterized in that, The operation of the calibrated UE trajectory filter in step S2 includes: A1: Particle initialization: Randomly initialize multiple particles near the initial position of the UE. Each particle represents a possible motion state of the UE, and randomly initialize the motion of the particles; A2: Particle propagation: According to the dynamic update rule of the motion information, update the motion information of each particle to simulate the motion of the UE; A3: Measurement update: According to the CSI measurement value provided by the known AP, calculate the predicted CSI of each particle, and compare it with the actual measurement value to update the weight of the particle; A4: Normalize weights: Normalize the weights of the particles to ensure that the sum of the weights of all particles is 1; A5: Resampling: Perform resampling to ensure uniform distribution of particles and avoid particle degeneracy; A6: Position estimation: Obtain the estimated motion state of the UE by weighted averaging the positions of all particles.

3. A multi-AP collaborative topology localization method based on particle filter according to claim 2, characterized in that In step A2, in order to simulate the real motion trajectory of the UE, a uniformly accelerated motion model is adopted in the X direction, a uniform linear motion is adopted in the Y direction, and the position and speed of the UE are updated. At the same time, simulate the motion data read by the IMU and add noise interference. The update formula is as follows: user_pos_false(t + 1)=user_pos_false(t)+user_vel_false(t)×T + 0.5×a meas ×T 2 where a meas is the acceleration measurement value after adding noise.

4. The multi-AP collaborative topology positioning method based on particle filter according to claim 3, characterized in that, In step A2, at each time step, the particle filter updates the speed and position of the particles through the motion model; the motion model is as follows: px_user(3,:) = px_user(3,:) + a particles × T px_user(1,:) = px_user(1,:) + px_user(3,:) * T + 0.5 * a particles * T 2 .

5. A multi-AP collaborative topology localization method based on particle filtering according to claim 4, characterized in that In step A3, update the weight of the particle: The likelihood function from the Gaussian distribution represents the joint likelihood of the independent observation errors of AP1 and AP2: pq_user = exp(-0.5*(vy_AP1 / v_std).^2).*exp(-0.5*(vy_AP2 / v_std).^2); where vy_AP1 and vy_AP2 are the differences between the true CSI of the user relative to the known AP and the CSI of the particle relative to the known AP; v_std is the standard deviation of the observation error.

6. The multi-AP collaborative topology localization method based on particle filter according to claim 5, wherein, The operation of the estimated AP position filter in step S3 includes: B1: Particle initialization: The estimated AP position filter randomly initializes particles within the possible position range of the unknown AP, and the particle positions are uniformly distributed within the preset area; B2: Particle propagation: The particles propagate according to the process noise to simulate the change of the unknown AP position; B3: Measurement update: According to the CSI measurement values between the unknown AP and the UE, calculate the predicted CSI of each particle, and compare it with the actual measurement value to update the weight of the particle; B4: Normalize weights: Normalize the weights of all particles to ensure that the sum of the weights of the particles is 1; B5: Resampling: Perform resampling to ensure that particles are concentrated in high-weight regions and avoid particle degeneracy; B6: Position estimation: Obtain the estimated position of the unknown AP by weighted averaging the positions of all particles.

7. A multi-AP collaborative topology localization method based on particle filtering according to claim 6, characterized in that Updating the weights of particles in step B3: pq_AP3 = exp(-0.5 * (vy_AP3_pf2 / v_std).^2); where vy_AP3_pf2 is the difference between the true CSI of AP3 relative to the user and the CSI of the particle relative to the user.

8. A multi-AP collaborative topology positioning system based on particle filter, characterized in that including: Multiple APs for providing CSI; An IMU for obtaining the motion information of the UE; A corrected UE trajectory filter for estimating the real-time position of the UE based on the motion information and CSI; An estimated AP position filter for estimating the position information of the unknown AP based on the UE position and the CSI of the unknown AP.

Images