Time of arrival positioning of non-line of sight anchor points based on channel knowledge maps

Abstract

The application discloses a non-line-of-sight anchor point screening method in time of arrival positioning based on a channel knowledge map. In a time of arrival (ToA) positioning system, non-line-of-sight (NLoS) links can cause positive time delay bias and reduce the positioning performance. The application uses a particle filtering method to collect particles uniformly on a channel knowledge map (CKM), and updates the particle weight according to the line-of-sight link information contained in the CKM to obtain a position estimate. Subsequently, the CKM is used to read out the anchor points corresponding to the position estimate, and a new position estimate is calculated according to the new anchor point set. The process is repeated, and the iteration between the anchor points and the position is performed until the anchor points converge. An effective means is added to avoid the anchor points from falling into deterioration in the iteration process, so that the screening ability of the positioning system for the NLoS anchor points is effectively improved, and the overall positioning performance is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of wireless positioning technology, and particularly relates to a channel knowledge map-based ToA positioning non-line-of-sight anchor point screening method. BACKGROUND

[0002] Since the emergence of the second generation (2G) mobile communication network, positioning with ground anchors has developed into one of the most important components of cellular networks. For the future sixth generation (6G) mobile communication network, it is believed that centimeter-level indoor positioning accuracy and sub-meter-level outdoor positioning accuracy should be achieved, but the positive random deviation of non-line-of-sight link propagation in wireless positioning causes a great influence on the positioning performance of ToA-based positioning. Currently, non-line-of-sight link errors can be reduced by two common methods: non-line-of-sight link identification and non-line-of-sight link mitigation. Non-line-of-sight link identification attempts to identify distance measurement values containing non-line-of-sight link errors and uses line-of-sight link anchors for positioning, and non-line-of-sight link mitigation techniques use all non-line-of-sight link and line-of-sight link measurements for positioning, but try to minimize the impact of NLoS anchors. Therefore, non-line-of-sight link identification is very suitable as a pre-step for non-line-of-sight link mitigation and greatly helps the positioning performance of ToA positioning. The current common non-line-of-sight link identification methods mainly have the following ways:

[0003] (1) Prior information-based scheme, that is, using the prior information of NLoS anchors and LoS anchors, the NLoS anchors and LoS anchors are judged by the method of Bayesian decision. This kind of method is simple in design, but it is difficult to obtain accurate prior information.

[0004] (2) Deep learning-based scheme, this kind of method usually uses deep learning to extract features of various signal parameters, and trains a neural network model to judge NLoS / LoS for different signals. This way has high accuracy in a specific scene, but needs to specially build and train a neural network, and has poor generalization ability, migration ability and explainability.

[0005] (3) Filter-based scheme, this kind of method uses Kalman filter or particle filter to constantly correct the positioning through time-varying information within a certain time to obtain better positioning performance. This kind of method is suitable for mobile positioning, but cannot obtain good positioning performance in one signal transmission.

[0006] (4) Abnormal detection-based scheme, this kind of method such as the least median square method uses a sample subset to calculate the residual, and selects the combination with the lowest residual index. This kind of method does not require any prior information, but the screening ability for NLoS anchors is not ideal.

[0007] The emergence of the channel knowledge map provides a new opportunity for NLoS anchor point removal in ToA positioning. The slow-varying channel information stored in the channel knowledge map can be more effectively and more targeted collected through online recording and offline measurement, and can play a role in other wireless communication fields such as beam alignment, without the need to increase the cost for the positioning system alone. But the slow-varying channel information stored in the channel knowledge map is anchored to the actual geographical location, and it is difficult to use the information when the user position information is not clear.

[0008] Based on the analysis of the above-mentioned NLoS anchor point removal scheme in ToA positioning, it can be seen that in the field of ToA positioning, there is an urgent need for an NLoS anchor point screening scheme that can effectively use the slow-varying channel information provided by the channel knowledge map, so as to improve the identification rate of NLoS link anchors and reduce the root mean square error of ToA positioning. SUMMARY

[0009] The present application aims to provide a channel knowledge map-based ToA positioning non-line-of-sight anchor point screening method to solve the problem that the existing method cannot use the slow-varying channel information provided by the channel knowledge map to screen NLoS anchors.

[0010] To solve the above technical problems, the specific technical solutions of the present application are as follows:

[0011] A channel knowledge map-based ToA positioning non-line-of-sight anchor point screening method, comprising the following steps:

[0012] Step 1, initialize particles and perform initial value estimation of user position: after obtaining the signal arrival time measurement data from each anchor point and the geographical position of each anchor point, uniformly sample on the map; take the position coordinates obtained by sampling as particle coordinates, and read out the LoS anchor point set corresponding to the particle coordinates from the LoS map, and calculate the weight of the particle according to the posterior residual value of the LoS anchor point set; then, take the coordinate value of the particle with the highest weight as the initial position estimate of the user; wherein the LoS map is a CKM containing the LoS link condition between each geographical position point on the map and each anchor point;

[0013] Step 2, verify each other using anchors and CKM: input the user position estimate at the current stage into the LoS map to obtain the LoS anchor point set corresponding to the position estimate, and judge whether the number of anchor points in the LoS anchor point set is greater than 3 to meet the requirement of calculating a new position estimate; when the subsequent iteration step returns to step 2, an iteration number judgment needs to be performed.

[0014] If the iteration number has exceeded the limit, the particle filter iteration is completed, and step 5 is entered; the maximum iteration number is the same as the maximum particle iteration number.

[0015] If the new position estimate cannot be calculated, step 4 is entered to update the particle;

[0016] If the new position estimate can be calculated, a new user position estimate is calculated according to the LoS anchor set by the least square method; then the LoS map is used to continue reading the LoS anchor set corresponding to the new user position estimate, and convergence is judged; if the LoS anchor set read by the LoS map for the new user position estimate is the same as the anchor set obtained by the least square method for the new user position estimate, the anchor converges, and the current new user position estimate is directly output; otherwise, step 3 is entered to delete the anchor;

[0017] Step 3, in the anchor deletion stage, if the anchor deletion stage has been performed in the last iteration, step 3 is skipped, and step 4 is executed to update the particle; otherwise, all abnormal anchors in the anchor set are deleted, and then the remaining anchors are used as the initial anchors for the next iteration, i.e., step 2 is entered;

[0018] Step 4, before updating the particle, the particle update number is first judged; if the particle update number has reached the threshold, step 5 is directly entered, otherwise, the particle is updated according to the LoS information provided by the LoS map;

[0019] Step 5, when the particle update number reaches the threshold, the particle is updated one last time and the user position estimate is obtained as the particle filter result; the anchor set corresponding to the user position estimate is read out by the LoS map, and another user position estimate is obtained from the anchor set; the reliability of the user position estimate calculated from the anchor is judged; if it is reliable, it is accepted as the final user position estimate, otherwise, the particle filter result is taken as the final user position estimate.

[0020] Further, in step 1, the LoS map stores slow-varying channel information of the line-of-sight link condition between all geographical position points and anchors on the map.

[0021] Further, in step 1, after reading out the line-of-sight anchor set corresponding to the particle coordinate point by the LoS map, the residual value of each coordinate point is calculated, and the reciprocal of the residual value is normalized as the weight of the particle.

[0022] Further, in step 3, the deletion of anchor points needs to calculate the root of the residual value between the current anchor point set and the user position estimation by using the LoS map, and calculate the mean and standard deviation of the root of the residual value to delete the anchor points in the anchor point set whose residual root is higher than the mean by one standard deviation; then the remaining anchor points after deletion continue to be deleted by the above method until no anchor point can be deleted, at which time the deletion step of the abnormal anchor point is completed.

[0023] Further, in step 4, the particle update needs to be performed according to the geographic information provided by the LoS map without receiving a new round of measurement data from the positioning anchor points; when updating the particles, first, the low-weight particles are deleted according to the average of the particle weights and the weights are normalized; then the remaining particles are used as the center value of the new particles, and the weight of the remaining particles is used as the probability of selecting the center value; after selecting the center value, a two-dimensional Gaussian noise with a mean of 0 is superimposed to improve the randomness of the particles, and one particle generation is completed; the above particle generation process is repeated until the number of particles is the same as before the particles are deleted.

[0024] Further, in step 5, the reliability of the user position estimation calculated by the anchor point depends on the distance between the user position estimation and the particle filtering result, and if the distance is greater than half of the sampling interval when the initial particles are generated, it is considered to be unreliable.

[0025] The method has the following advantages:

[0026] 1. The method does not need to use any prior information other than the channel knowledge map, and does not need to perform additional prior information measurement for positioning in a system equipped with a channel knowledge map. In step 1, the user position prior information is not used for initialization; in step 4, new positioning signal measurement is not needed to update the particles.

[0027] 2. The method effectively solves the problem of how to use the channel information provided by the channel knowledge map to screen out NLoS anchor points when the user position is unknown. The uniform sampling in step 1 and the iteration of anchor points and position estimation in step 2 make the channel knowledge map in this method independent of the exact position of the user.

[0028] 3. The method can fully utilize the channel information provided by the channel knowledge map, and effectively improve the recognition rate of NLoS anchor points without using other prior information. The anchor point deletion in step 3 greatly reduces the probability of NLoS anchor points in the final recognition result, and step 2 can find as many missed LoS anchor points as possible. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1is a wireless positioning method based on time of arrival in urban environment provided by the embodiment of the present application;

[0030] Figure 2 is a method flow diagram for NLoS anchor point screening based on channel knowledge map provided by the embodiment of the present application;

[0031] Figure 3 is a simulation scene diagram provided by the embodiment of the present application;

[0032] Figure 4 is a simulation performance diagram about NLoS anchor point clearance success rate provided by the embodiment of the present application;

[0033] Figure 5 is a simulation performance diagram about root mean square error of positioning result provided by the embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to better understand the purpose, structure and function of the present application, the present application will be further described in detail below in combination with the drawings.

[0035] The embodiment of the present application provides a non-line of sight anchor point screening method in ToA positioning based on CKM. Through the line of sight link state information stored in CKM, the mutual verification between anchor points and user position estimation is realized, and the problem of how to use CKM without specific user position is solved. Meanwhile, the NLoS anchor point screening method based on CKM can effectively improve the identification ability of NLoS anchor points and reduce the root mean square error of positioning without using other prior information.

[0036] In one embodiment, referring to the flow diagram shown in Figure 2 The present application provides a non-line of sight anchor point screening method in ToA positioning based on CKM, which includes the following main steps:

[0037] a) Initialize particles and perform initial value estimation of user position. Specifically, after obtaining the signal arrival time measurement data from each anchor point and the geographical position of each anchor point, uniform sampling is performed on the map. The position coordinates obtained by sampling are taken as particle coordinates, and the LoS anchor point set corresponding to the particle coordinates is read out by using the LoS map, and the weight of the particle is calculated according to the residual error of the LoS anchor point set; subsequently, the coordinate value of the particle with the highest weight is taken as the initial position estimation of the user. The LoS map is a CKM containing the line of sight link state between each geographical position point on the map and each anchor point.

[0038] b) Mutual verification using anchor points and channel knowledge map. Specifically, the user position estimate of the current stage is input into the LoS map to obtain the LoS anchor point set corresponding to the position estimate, and it is judged whether the number of anchor points in the LoS anchor point set is greater than 3 to meet the requirement of calculating a new position estimate. If a new position estimate cannot be calculated, step d) is entered to update the particle; if a new position estimate can be calculated, the LoS anchor point set is used to calculate the new user position estimate by the least square method; then the LoS map is used to continue reading the LoS anchor point set corresponding to the new user position estimate, and a convergence judgment is made. If the LoS anchor point set read out by the LoS map for the new user position estimate is the same as the anchor point set obtained by the least square method for the new user position estimate, the anchor points converge, and the current new user position estimate is directly output; otherwise, step c) is entered to delete the anchor points. When the subsequent iteration step returns to step b), an iteration number judgment needs to be made. The maximum number of iterations is the same as the maximum number of iterations of the particle. If the number of iterations has exceeded the limit, the number of iterations required for particle filtering is supplemented, and step e) is entered.

[0039] c) In the anchor point deletion stage, if the anchor point deletion stage has been performed in the last iteration, step c) is skipped, and step d) is executed to update the particle; otherwise, all abnormal anchor points in the anchor point set are deleted, and then the remaining anchor points are used as the initial anchor points for the next iteration, i.e., returning to step b).

[0040] d) Before updating the particle, a particle update number judgment is first made. If the particle update number has reached the threshold, step e) is directly entered; otherwise, the particle is updated according to the LoS map provided line-of-sight information.

[0041] e) When the particle update number reaches the threshold, the particle is updated one last time and the user position estimate is obtained as the particle filtering result; the anchor point set corresponding to the user position estimate is read out using the LoS map, and another user position estimate is obtained from the anchor point set; the user position estimate calculated from the anchor points is judged for reliability. If it is reliable, it is accepted as the final user position estimate; otherwise, the particle filtering result is taken as the final user position estimate.

[0042] wherein,

[0043] In step a), the channel knowledge map used is the LoS map, which stores slow-varying channel information of line-of-sight link conditions between each geographic location and each anchor point, and the particle is a set of sampling coordinate points with weights. The particle sampling interval needs to be flexibly selected according to different geographic environments. After reading out the line-of-sight anchor point set corresponding to the coordinate point using the LoS map, the residual value between the measured distance and the distance between each line-of-sight anchor point and each coordinate point needs to be calculated, and then the reciprocal of the residual value of each coordinate point is normalized to serve as the weight of the particle.

[0044] In step c), the root of the residual between the current anchor set and the user position estimate is calculated using the LoS map, and the mean and standard deviation of the set of residual roots are calculated to remove anchors whose residual root is higher than the mean by more than one standard deviation. The anchor removal step is then continued with the remaining anchors until no anchors can be removed, at which point the anchor removal step is complete.

[0045] In step d), the particles are updated based on the geographical information provided by the LoS map without receiving a new round of measurement data from the positioning anchors. In the particle update, low-weight particles are first removed based on the mean of the particle weights, and the weights of the particles are then normalized. The remaining particles are then used as the optional center values of the new particles, and a discrete probability distribution of the center values of the new particles is established based on the weights of the remaining particles, i.e., the probability of selecting a center value is equal to the weight of the particle corresponding to the center value. After selecting the center value, a 0-mean two-dimensional Gaussian noise is superimposed on the center value as a random disturbance to improve the randomness of the particles, and thus one particle generation is completed. The standard deviation of the two-dimensional Gaussian noise should not be greater than the sampling rate of the particles. The above particle generation process is repeated until the number of particles is the same as before the particles are removed. In this update process, the convergence speed of the particles is fast, and thus a long time of particle iteration is not required.

[0046] In step e), the reliability of the user position estimate calculated by the anchors is determined based on the distance between the user position estimate and the particle filtering result. If the distance is greater than half of the sampling interval when the initial particles are generated, it is considered to be unreliable.

[0047] The method is not only suitable for single-user positioning scenarios, but also for multi-user positioning scenarios. For multi-user positioning scenarios, either a distributed scheme can be used to load the channel knowledge map by each positioning terminal for screening, or a centralized scheme can be used for unified processing.

[0048] Figure 1 is a schematic diagram of a wireless positioning based on time of arrival in an urban environment according to an example embodiment, where r i represents the distance measurement calculated by the i-th anchor by time of arrival. Therefore, when the anchor is a LoS anchor, the measurement value r i is determined by the actual distance d i between the anchor and the user, and the 0-mean Gaussian noise n i superimposed thereon, i.e., r i = d i + n i ; if the anchor is a non-line-of-sight anchor, a positive random bias b i is also superimposed, i.e., ri =d i +n i +b i Therefore, the residual value res of a certain estimate can be defined as... in, This represents the set of Loss anchors used to obtain the estimated value. This represents the number of anchor points in the set of anchor points. This is the distance between the user and the i anchor points calculated based on the current estimated location. When updating the particle filter to calculate the residual, Use the set of anchor points obtained from the Loss map as a substitute.

[0049] Based on the above description and definitions, the specific implementation steps of the exemplary embodiment of the proposed method can be summarized as follows:

[0050] (1) Initialization phase. Within the map service area of ​​the anchor point, n coordinate points p are uniformly selected in two-dimensional space at a distance D, and the resulting n coordinate points p are... n Send to Los map Obtain the set of sight distance anchor points corresponding to this coordinate point. And calculate the coordinates of each point p. n Residual value Res n Therefore, the weight w of each particle n It can be represented as That is, a particle can be represented as (p n ,w n Then, the particle with the highest weight is used as the initial user estimate p0, and the process is iterated.

[0051] (2) Estimation and Anchor Point Iteration Phase. The user estimate p in the current iteration phase... s Inputting the Loss map yields the corresponding set of Loss anchor points. and utilize Calculate the new user estimate p s+1 And read out the user's estimated value. like If convergence is achieved, then the process is complete; otherwise, the process proceeds to the anchor point deletion phase. In this phase, if the anchor point deletion phase has already been entered in the previous iteration, or if the number of anchor points in a certain anchor point set cannot provide a user position estimate, then the process directly proceeds to the particle filter update phase. If the number of iterations has reached a set threshold, then particle filtering is performed several times until the number of particle filter updates reaches the set threshold T, and then the process proceeds to the particle filter output phase.

[0052] (3) Anchor point deletion stage. The new position estimate p is obtained. s+1 Anchor point set Calculate the root residual value of each anchor point in the equation. And calculate the mean value μ of the square roots of the residuals in this group sRes and the standard deviation σ sRes . Subsequently, delete all the anchor points where sRes - μ sRes > σ sRes to obtain a new set of anchor points Regard as Repeat the above anchor point deletion process until there are no anchor points that can be deleted in the newly obtained set of anchor points. At this time, the anchor point deletion process ends. Finally, calculate a new user position estimate using the set of anchor points after deletion and return to the estimation and anchor point iteration stage. If the number of anchor points after deletion is not sufficient to calculate a new position estimate, enter the particle filter update stage.

[0053] (4) Particle filter update stage. Calculate the mean value of all particle weights, delete all particles with weights lower than this mean value, and normalize the weights of the remaining particles p' n to obtain new weights w' n . Subsequently, according to the weights of the remaining particles, use the remaining particles as the value range of the center values of the newborn particles, that is, the center values of the newborn particles p c obeys the probability distribution function Pr{p c = p' n} = w' n of a discrete random distribution. Then generate the coordinate values p” n of the new particles, and the coordinates of the newborn particles satisfy p” n = p c + n p , until the number of particle coordinates returns to before the update, where n p is a two-dimensional zero-mean Gaussian random variable with a standard deviation of σ p , that is where σ p < D. Finally, assign values to the particle coordinates referring to the initialization stage, complete the particle update, and obtain a new user position estimate and return to the estimation and anchor point iteration stage. If the number of executions of the particle filter update stage has reached the set threshold T, enter the particle filter output stage.

[0054] (5) Particle filter output stage. Finally, perform a particle filter update once and obtain the user position estimate p par corresponding to the particle filter. Subsequently, use the LoS map to read out the set of anchor points par corresponding to the user position estimate p and use this set of anchor points to obtain another user position estimate p anc . Calculate the distance dis = ||p anc - p par2. If dis > D / 2, then the particle filter result p par Use it as the end-user location estimate. Otherwise, use the anchor point estimate result p anc As an end-user location estimate.

[0055] Example Performance Simulation

[0056] Figure 3 This is a schematic diagram of the scenario used for performance simulation in this embodiment. Several obstacles and 10 randomly located positioning anchor points are placed within an area 1 km wide and long. The non-line-of-sight path deviation value is a positive random value of 0.25-1 times the distance between the user and the anchor point. The user's position is randomly generated within the area within the dashed box in the schematic diagram, but not in locations where the number of available LoS anchor points is less than 3. 1000 user position points meeting the requirements described above are selected, such as... Figure 4 and Figure 5 As shown, the proposed method and the improved least squares median method are plotted respectively. [1] The ability to clear NLoS anchor points and the variation of the root mean square error of positioning with different Gaussian noise standard deviations. Figure 4 The NLoS clearance success rate is defined as the proportion of location points in the selected anchor point set that do not contain NLoS anchor points out of all generated user location points after NLoS identification. Figure 5 The definition of the root mean square error is Where p j For the j-th generated user location point, Let be the estimated value of the j-th generated user location point. It is evident that the method proposed in this invention significantly improves performance compared to benchmark methods, both in terms of NLoS anchor point identification capability and the overall root mean square error of the positioning system.

[0057] in [1] The reference is "T.Qiao and H.Liu, "Improved Least Median of SquaresLocalization for Non-Line-of-Sight Mitigation," in IEEE Communications Letters, vol.18, no.8, pp.1451-1454, Aug.2014".

[0058] In summary, by using a Loss map, this invention solves the problem of filtering out NLoS anchor points for ToA positioning using slowly varying channel state information in a channel knowledge map when there is no specific user location, without using other prior information. This effectively improves the positioning system's ability to filter NLoS anchor points and reduces the root mean square error of ToA positioning.

[0059] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A channel knowledge map based time of arrival positioning non-line of sight anchor point screening method, characterized in that, The method comprises the following steps: Step 1, initializing particles and performing initial value estimation of user position: after obtaining signal arrival time measurement data from each anchor point and the geographic position of each anchor point, uniformly sampling on the map; taking the position coordinates obtained by sampling as particle coordinates, reading out the LoS anchor point set corresponding to the particle coordinates by using the LoS map, and calculating the weight of the particle according to the posterior residual value of the LoS anchor point set; subsequently, taking the coordinate value of the particle with the highest weight as the initial position estimation of the user; wherein the LoS map is a CKM containing the LoS link condition between each geographic position point on the map and each anchor point; Step 2, verifying each other by using the anchor point and the CKM: inputting the user position estimation at the present stage into the LoS map to obtain the LoS anchor point set corresponding to the position estimation, and judging whether the number of anchor points in the LoS anchor point set is greater than 3 to meet the requirement of calculating a new position estimation; when the subsequent iteration step returns to step 2, it is necessary to judge the iteration number of times; If the iteration number of times has exceeded the limit, the iteration number of times required by the particle filter is supplemented, and step 5 is entered; the maximum iteration number of times is the same as the maximum iteration number of times of the particle; If the new position estimation cannot be calculated, step 4 is entered to update the particle; If the new position estimation can be calculated, the new user position estimation is calculated by using the least square method according to the LoS anchor point set; subsequently, the LoS anchor point set corresponding to the new user position estimation is read out by using the LoS map, and convergence judgment is performed; if the LoS anchor point set read out by the LoS map from the new user position estimation is the same as the anchor point set obtained by the least square method from the new user position estimation, the anchor points converge, and the new user position estimation at the present stage is directly output; otherwise, step 3 is entered to delete the anchor points; Step 3, in the anchor point deletion stage, if the anchor point deletion stage has been performed in the last iteration, step 3 is skipped, and step 4 is executed to update the particle; otherwise, all abnormal anchor points in the anchor point set are deleted, and subsequently, the remaining anchor points are used as the initial anchor points for the next iteration, that is, step 2 is entered; in step 3, the root of the residual value between the present anchor point set and the user position estimation is calculated by using the LoS map, and the mean and standard deviation of the root of the residual value are calculated to delete the anchor points in the anchor point set whose residual root is higher than the mean by one standard deviation; subsequently, the remaining anchor points after deletion are continuously subjected to anchor point deletion until no anchor point can be deleted, and the step of deleting the abnormal anchor points is completed; Step 4, before updating the particle, the particle update number of times is first judged; if the particle update number of times has reached the threshold, step 5 is directly entered; otherwise, the particle is updated according to the LoS information provided by the LoS map; Step 5, when the number of particle updates reaches a threshold, the last particle update is performed and a user position estimate is obtained as the particle filter result; the anchor set corresponding to the user position estimate is read out from the LoS map, and another user position estimate is obtained from the anchor set; the user position estimate obtained from the anchor set is judged for reliability, and if reliable, it is accepted as the final user position estimate, otherwise, the particle filter result is accepted as the final user position estimate.

2. The channel knowledge map based time of arrival positioning non-line of sight anchor screening method of claim 1, wherein, In step 1, the LoS map stores slow-varying channel information of the LoS link between all geographical position points on the map and each anchor.

3. The channel knowledge map based time of arrival positioning non-line of sight anchor screening method of claim 1, wherein, In step 1, after reading out the LoS anchor set corresponding to the particle coordinate point from the LoS map, the residual value of each coordinate point needs to be calculated, and the reciprocal of the residual value is normalized as the weight of the particle.

4. The channel knowledge map based time of arrival positioning non-line of sight anchor screening method of claim 1, wherein, In step 4, the particle update needs to be performed according to the geographical information provided by the LoS map without receiving a new round of measurement data from the positioning anchor; during the particle update, low-weight particles are first deleted according to the average value of the particle weights and the weights are normalized; then the remaining particles are used as the center value of the new particles, and the weight of the remaining particles is used as the probability of selecting the center value; after selecting the center value, a two-dimensional Gaussian noise with a mean value of 0 needs to be superimposed on the center value to improve the randomness of the particles, and one particle generation is completed; repeat the particle generation process until the number of particles is the same as before the particles are deleted.

5. The channel knowledge map based time of arrival positioning non-line of sight anchor screening method of claim 1, wherein, In step 5, the reliability of the user position estimate obtained from the anchor set depends on the distance between the user position estimate and the particle filter result, and if the distance is greater than half of the sampling interval when the initial particle is generated, it is considered to be unreliable.

Images