Multi-path amplitude-independent robust snapshot radio frequency SLAM (Simultaneous Localization and Mapping) method

Abstract

The invention provides a multipath amplitude-independent robust snapshot radio frequency SLAM (Simultaneous Localization and Mapping) method, and belongs to the field of synchronous localization and mapping of radio signals. Positioning and environment mapping are carried out by analyzing channel parameters extracted by radio signals from a base station and a user, and a single snapshot geometric measurement model with independent amplitudes is firstly constructed; and estimating the position and direction of the user based on a single snapshot geometric measurement model with independent amplitudes. According to the method, the accuracy of user position and direction estimation is improved, and the problem that the process of estimating the user position and direction by using the radio frequency SLAM depends on amplitude information is solved, so that the SLAM system can provide high-precision user position and direction estimation under a non-line-of-sight condition.

Description

Technical fields:

[0001] This invention belongs to the field of synchronous positioning and mapping of radio signals, and particularly relates to a robust snapshot radio frequency SLAM method with multipath amplitude independence. Background technology:

[0002] In wireless networks, localization and awareness based on downlink / uplink signals between the base station (BS) and the user equipment (UE) typically employ a two-stage process: The first stage involves estimating channel parameters such as angle of arrival (Angle of Arrival), departure angle (Angle of Departure), and time of arrival (Time of Arrival) along the propagation path; the second stage utilizes these channel parameters for localization and environmental awareness. Resolvable non-line-of-sight (NLoS) paths not only provide the UE's location information but also offer characteristic information about single-reflection NLoS paths. Therefore, the solvability of channel parameters between the base station and the UE enables radio SLAM technology to simultaneously perform localization and environmental mapping, estimating not only the UE's state but also simultaneously constructing an environmental map.

[0003] The prior art disclosed in publication number CN112415522A is a semantic SLAM method for real-time perception of dynamic object information, relating to the field of SLAM dynamic perception technology. This semantic SLAM method for real-time perception of dynamic object information includes the following steps: First, constructing a 3D global coordinate system on a map and marking the initial position of the dynamic object to be perceived in the coordinate system on the map; Second, installing a laser rangefinder and an ultrasonic rangefinder on the surface of the dynamic object to be perceived, both of which are connected to a central processing unit (CPU) via a wireless signal conversion and transmission component; Third, connecting the CPU to the computer that established the global coordinate system. However, this method requires high computing power and has a long computation time, making it unsuitable for many scenarios.

[0004] In radio SLAM, one of the core challenges in localization and environment mapping using multiple reflection paths is distinguishing between Loss-of-Sight (LoS) and Non-LoS (NLoS) paths. This involves processing numerous channel parameters, which in turn affects system performance. This approach enables accurate localization and environmental awareness even in the absence of a line-of-sight path. The core of radio SLAM systems lies in optimizing localization performance by analyzing the angle and time-of-arrival information of the propagation path, especially in complex multipath and low-signal environments. Existing radio SLAM methods typically rely on filtering or snapshot methods for path estimation. Filtering methods, by recursively estimating the UE state and environment map, solve the problem of continuous localization by processing observation data and generally perform well in multipath propagation environments. However, this method has high computational complexity, especially when processing large amounts of observation data. Snapshot SLAM methods typically solve single-shot localization tasks without requiring motion models or prior information, making them suitable for certain specific scenarios. However, traditional snapshot methods also have limitations due to their reliance on path detection and signal amplitude models, particularly in noisy and multipath propagation environments, where their localization accuracy significantly decreases. Therefore, finding a SLAM optimization method that addresses these issues is crucial. Summary of the Invention:

[0005] Objective: To address the shortcomings of existing technologies, this invention designs a robust snapshot RF SLAM method independent of multipath amplitude. It achieves accurate estimation of UE position and state through geometric constraints, avoiding the dependence on signal amplitude models in traditional methods, and effectively distinguishes between LosS and NLoS paths using path angle information. To optimize LosS path detection, this invention also introduces an AIC-based model selection strategy, thereby further improving the system's robustness and accuracy in multipath environments.

[0006] Technical Solution: This invention discloses a multipath amplitude-independent robust snapshot radio frequency SLAM method for use in a system formed by a base station and a user, comprising the following steps:

[0007] S1. Using the base station location, user location, scattering point location, and the departure angle AoD and arrival angle AoA information of the signal propagation between the base station, user and scattering point, construct an amplitude-independent single snapshot geometric measurement model. In the single snapshot geometric measurement model, geometric mapping is used to form a model based only on the line-of-sight path LoS and the single non-line-of-sight path NLoS-1.

[0008] S2. Based on the single snapshot geometric measurement model, the user position, system clock deviation and user orientation information are jointly estimated and then subjected to primary joint estimation and robust joint estimation to realize snapshot radio frequency SLAM.

[0009] S2.1 Initial Joint Estimation: Using a single snapshot geometric measurement model, based solely on the channel parameter information of the line-of-sight path LoS and the single non-line-of-sight path NLoS-1, an initial joint estimation of user location, system clock offset, and user direction is performed.

[0010] S2.2 Robust Joint Estimation: Based on the aforementioned primary joint estimation method, measurement information including multiple scattering paths is further introduced; using the RANSAC algorithm, a valid subset of measurements contributed by line-of-sight and single scattering paths is robustly selected from the complete measurement set containing line-of-sight Loss, single scattering NLoS-1, and multiple scattering paths; based on the selected valid subset of measurements, a final robust joint estimation of user position, system clock offset, and user orientation is performed.

[0011] Furthermore, the process of constructing an amplitude-independent single-snapshot geometric measurement model is as follows:

[0012] On the XYZ spatial coordinate axis, the single scattering volume point p formed by the intersection of the i-th single scattering path. i With the coordinates of the point p as the center, and the single scattering point p as the center. i Distance from base station coordinates p BS and user coordinates p UE Construct two concentric circles with radius p. i To user coordinates p UE The distance and p i to base station coordinates p BS The distance, line segment (p) i p UE The radius of ) is smaller than that of line segment (p) i p BS The radius of p BS and p UE Located on two circles respectively, p BS to p i The extensions of the line segments intersect p UE The focus of the circle is point E, p UE to p i The extensions of the line segments intersect p BS Let point A be the focus of the circle, and let p... BS to p i The unit vector is denoted as u i That is, the direction vector of AOD, p i to p UE The unit vector is denoted as v i That is, the direction vector of AOA, and the vector containing the line segment (A, E) is denoted as vector m. i To simplify the calculation, the result of the cross product of vectors in three-dimensional space is projected onto the plane of the XY axis, that is, the Z-axis component is set to 0 for calculation.

[0013] p BS With p i and p UE The included angle φ formed by the lines connecting them i -φ LoS With point A and p i The angle φ formed by the lines connecting point E and point E i ' are equal, where φ LoS This represents the departure angle of the line-of-sight path, therefore vector m i and v i The cross product of vector u is equal to that of vector u. i The cross product of t and t, where t = p UE -p BS The geometric relationship between the coordinate difference between the base station and the user satisfies:

[0014] d i v i ×m i =d i u i ×t

[0015] Where d i =c(τ) i -b UE ) represents the propagation distance of the i-th path, τ i Let c represent the propagation time of the i-th single path, and c represent the speed of light.

[0016] Based on the geometric relationship of concentric circles, m i Rewritten as:

[0017] m i =d i u i -(t+d i v i )

[0018] Where, m i This represents the vector containing the line segment AE from point A to point E;

[0019] d i v i ×m i =d i u i The ×t rearrangement constructs an amplitude-independent single-snapshot geometric measurement model under noise-free conditions as follows:

[0020] d i (u i ×v i )=t×(u i +v i ).

[0021] Furthermore, the primary joint estimation method in step S2.1 is as follows:

[0022] A single snapshot geometric measurement model based on noise-free conditions with independent amplitude. i (u i ×v i )=t×(u i +v i Introducing residual r under noisy conditions i (x UE ):

[0023]

[0024] in Using the least squares (LS) method to obtain information about x UE The optimal estimate minimizes the least squares criterion of the residuals; when considering the measurements Z = [z1, z2, ..., z] of all base station to user propagation paths. n When using the likelihood function p(z) of the channel parameters... i ∣Θ i )=N(z i |h i (Θ i ),Σ i ), by analyzing the mean h i (Θ i Apply noise Σ i Multiple sets of measurements Z can be generated; for all paths, the following values ​​about x are constructed. UE Cost function:

[0025]

[0026] The optimal estimate is:

[0027]

[0028] Where I = [1,2,...,n] represents the path index value of the measurement Z for all n paths.

[0029] Furthermore, the robust joint estimation method in step S2.2 is as follows:

[0030] Suppose the index of the measurement value Z for n paths is... By finding a Z subset that does not contain information about multiple scattering paths, and then using the Z subset to estimate x UE ; Use the RANSAC algorithm to randomly select N from the measured values ​​Z min The total number of combinations L, which combines paths, is expressed as:

[0031]

[0032] in, This represents the index value of the path in all combinations. Nmin represents the minimum number of paths required to solve for x;

[0033] Let l be a combination within L, and let the path index corresponding to the l-th combination be denoted as . For the l-th combination A path, defined by the formula: Solve for the estimated value of the l-th combination;

[0034] Due to the path index of the measured value Z Including line-of-sight, single scattering path, and multiple scattering path, using the formula right The division is as follows:

[0035]

[0036] in Indicates in The Loss of Scattering (LoS) paths and single scattering paths, or only single scattering paths, are divided into interior points, J. i (x)≥T ε The path whose cost function is greater than the threshold is a multiple scattering path. The path is denoted as the multiple scattering path information, called the outgoing point, and denoted as the multiple scattering path. express Within the line-of-sight distance and single scattering path;

[0037] based on Recalculation passed The cost function C of the l-th combination is obtained. l :

[0038]

[0039] Where T ε C represents the detection threshold for multiple scattering paths. inliers C represents the cost function corresponding to line-of-sight distance and single scattering path. outliers This represents the cost function corresponding to multiple scattering paths;

[0040] The combination of minimum cost functions is labeled as The path index of the combination corresponding to the minimum cost function is denoted as Based on this, it was passed again The user's location, clock offset, and orientation are estimated to obtain the user's UE state.

[0041] Furthermore, based on the robustly jointly estimated UE state values ​​described above, a more refined estimate of the user UE state is obtained through iterative optimization using the Gauss-Newton algorithm, as detailed below:

[0042]

[0043] Θ = argmin Θ E(Θ)

[0044] Where E(Θ) represents the fitting error during the iteration process of the Gauss-Newton algorithm, z k Let Θ represent the measurement value of the k-th path, and let Θ represent the parameter value to be iteratively optimized, defined as... in express All paths except Loss. By minimizing E(Θ), the parameter value Θ to be optimized can be obtained;

[0045] LoS path parameter values For the parameter values ​​of the k-th NLoS path Therefore, it satisfies h k The parameterization of the measurement value of the k-th path is represented as follows:

[0046] If the k-th path is a Loss-free path:

[0047]

[0048] Where p BS and p UE These are the location coordinates of the base station and the user, respectively, α BS , α UE and b UE These are the clock offsets in the base station direction, the user direction, and the system direction, respectively. and Let c be the transpose of the matrix, and c be the speed of light.

[0049] If the k-th path is an NLoS path:

[0050]

[0051] Where p k The coordinates of the k-th path scattering point are given; the other parameters are the same as those for the line-of-sight path.

[0052] Furthermore, the process of identifying LoS and NLoS-1 paths is as follows:

[0053] Let the interior points be identified. There are M paths in total. Since each of these M paths between the base station (BS) and the user (UE) could potentially be a Loss-of-Stake (LoS) path, we construct M+1 candidate models for each LoS path, where M represents the number of LoS cases and 1 represents NLoS cases. For each of the M LoS cases, there is one path that uses h... LoS (Θ LoS ) to parameterize the measured values, using The fitting process yielded M+1 sets of error values ​​E(Θ), based on the Akaike Information Criterion (AIC) strategy:

[0054]

[0055] AIC m Θ represents the AIC value of the m-th model fit. (m) E represents the parameter value estimated by the m-th model. m The fitting error of the m-th model, |Θ (m) | represents the dimension of the m-th model. Let m represent the number of interior points, where m ∈ M;

[0056] Find the smallest AIC value among M+1 AIC values. The candidate model corresponding to the smallest AIC value clearly identifies the LosS and NLoS-1 paths.

[0057] A computer device includes a processor and a memory, the processor being electrically connected to the memory for storing instructions and data, and the processor for executing the multipath amplitude-independent robust snapshot RF SLAM method.

[0058] A computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed by the multipath amplitude-independent robust snapshot RF SLAM method.

[0059] Beneficial Effects: The method of this invention has significant technical advantages and practical application value. Through angular geometric constraints and a cross-product consistency mechanism, this method effectively improves positioning accuracy and system robustness in multipath propagation and high-noise environments, significantly outperforming traditional SLAM algorithms that rely on amplitude information. Employing a snapshot-style angle constraint strategy and an AIC model selection method, it can more accurately distinguish between LosS and NLoS paths, achieving efficient path classification and precise positioning. Furthermore, it does not rely on prior motion information or environmental models, making it suitable for dynamic, complex, or unknown scenarios. This method is widely applicable to fields requiring high-precision positioning, such as 5G / 6G communication, autonomous driving, industrial positioning, and underground navigation, demonstrating high practicality and widespread application potential. Attached image description:

[0060] Figure 1 This is a schematic diagram of a single snapshot geometric measurement model in an embodiment of the present invention. Detailed implementation method:

[0061] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0062] like Figure 1 The present invention discloses a multipath amplitude-independent robust snapshot radio frequency SLAM method for use in a system formed by communication between a base station (BS) and a user user (UE), wherein the BS state is determined by location. and direction α BS Represented by ∈[0,2π]. Through position p i Let represent the i-th scattering point, and let represent the intersection of the i-th single scattering paths. This point is called the single scattering volume point, and this path is denoted as the NLoS-1 path. UE state x UE Use position p UE Direction α UE and clock deviation b UE The following describes the specific steps for constructing a single geometric measurement model and solving for the UE state.

[0063] The specific steps are as follows:

[0064] Step 1. Construct a geometric measurement model, let z i =[τ i ,φ i ,θ i ] represents the estimated channel parameters for the i-th propagation path, also known as the observation, and let Z and Let represent a set of measurements and the associated path index, respectively. Assuming the measurement noise follows a zero-mean Gaussian distribution (a common assumption in SLAM), the likelihood function is also Gaussian distributed.

[0065] p(z i |Θ i )=N(z i |h i (Θ i ),Σ i )

[0066] Where h i (Θ i ) is the average value, Σ i It is covariance.

[0067] The following sections describe the construction of the observed mean values ​​for LoS and NLoS-1 paths. The observed mean values ​​for LoS and NLoS-1 paths are as follows:

[0068]

[0069] Where pBS and p UE These are the location coordinates of the base station and the user, respectively, α BS α UE and b UE These are the clock offsets in the base station direction, the user direction, and the system direction, respectively. and Let c be the transpose of the matrix, and c be the speed of light.

[0070]

[0071] The above equation represents the geometric relationship between the BS, UE, and scatterer. Where p i The coordinates of the i-th path scattering point are given, and the other parameters are the same as those for the line-of-sight path.

[0072] On the XYZ spatial coordinate axis, the single scattering volume point p is formed by the intersection of the i-th single scattering path. i With the coordinates of the point p as the center, and the single scattering point p as the center. i Distance from base station coordinates p BS and user coordinates p UE Construct two concentric circles with radius p. i To user coordinates p UE The distance and p i to base station coordinates p BS The distance, line segment (p) i p UE The radius of ) is smaller than that of line segment (p) i p BS The radius of p BS and p UE Located on two circles respectively, p BS to p i The extensions of the line segments intersect p UE The focus of the circle is point E, p UE to p i The extensions of the line segments intersect p BS Let point A be the focus of the circle, and let p... BS to p i The unit vector is denoted as u i That is, the direction vector of AOD, p i to p UE The unit vector is denoted as v i That is, the direction vector of AOA, and the vector containing the line segment (A, E) is denoted as vector m. i To simplify the calculation, the result of the cross product of vectors in three-dimensional space is projected onto the plane of the XY axis, that is, the Z-axis component is set to 0 for calculation.

[0073] p BS With p i and pUE The included angle φ formed by the lines connecting them i -φ LoS With point A and p i The angle φ formed by the lines connecting point E and point E i ' are equal, where φ LoS This represents the departure angle of the line-of-sight path, therefore vector m i and v i The cross product of vector u is equal to that of vector u. i The cross product of t and t, where t = p UE -p BS The geometric relationship between the coordinate difference between the base station and the user satisfies:

[0074] d i v i ×m i =d i u i ×t

[0075] Where d i =c(τ) i -b UE ) represents the propagation distance of the i-th path, τ i Let c represent the propagation time of the i-th single path, and c represent the speed of light.

[0076] Since the BS and UE have their own orientations, and their local coordinate systems are not unified, for ease of description, the coordinate axes are represented uniformly, and all relevant vectors are represented in the global coordinate system.

[0077] v i =R(α) UE )[cos(θ i sin(θ) i )] T

[0078] u i =R(α) BS )[cos(φ i sin(φ) i )] TT

[0079] Where R(α) is the counterclockwise rotation matrix.

[0080] Based on the geometric relationship of concentric circles, m i Rewritten as:

[0081] m i =d i u i -(t+d i v i )

[0082] Where, m i This represents the vector containing the line segment AE from point A to point E;

[0083] d i v i ×m i =d i u i The ×t rearrangement constructs an amplitude-independent single-snapshot geometric measurement model under noise-free conditions as follows:

[0084] d i (u i ×v i )=t×(u i +v i ).

[0085] Step 2. Amplitude-independent single snapshot geometric measurement model d under noise-free conditions i (u i ×v i )=t×(u i +v i Introducing residual r under noisy conditions i (x UE ):

[0086]

[0087] in Using the least squares (LS) method to obtain information about x UE The optimal estimate minimizes the least squares criterion of the residuals; when considering the measurements Z = [z1, z2, ..., z] of all base station to user propagation paths. n When using the likelihood function p(z) of the channel parameters... i |Θ i )=N(z i |h i (Θ i ),Σ i ), by analyzing the mean h i (Θ i Apply noise Σ i Multiple sets of measurements Z can be generated; for all paths, the following values ​​about x are constructed. UE Cost function:

[0088]

[0089] The optimal estimate is:

[0090]

[0091] Where I = [1,2,...,n] represents the path index value of the measurement Z for all n paths.

[0092] Step 3. Assume the index of the measured value Z for the n paths is... By finding a Z subset that does not contain information about multiple scattering paths, and then using the Z subset to estimate x UE ; Use the RANSAC algorithm to randomly select N from the measured values ​​Z min The total number of combinations L, which combines paths, is expressed as:

[0093]

[0094] in, Nmin represents the index value of the path in all combinations, and Nmin represents the minimum number of paths required to solve for x.

[0095] Let l be a combination within L, and let the path index corresponding to the l-th combination be denoted as . For the l-th combination A path, defined by the formula: Solve for the estimated value of the l-th combination;

[0096] Due to the path index of the measured value Z Including line-of-sight, single scattering path, and multiple scattering path, using the formula

[0097] right

[0098] The division is as follows:

[0099]

[0100] in Indicates in The Loss of Scattering (LoS) paths and single scattering paths, or only single scattering paths, are divided into interior points, J. i (x)≥T ε The path whose cost function is greater than the threshold is a multiple scattering path. The path is denoted as the multiple scattering path information, called the outgoing point, and denoted as the multiple scattering path. express Within the line-of-sight distance and single scattering path;

[0101] based on Recalculation passed The cost function C of the l-th combination is obtained. l :

[0102]

[0103] Where T ε C represents the detection threshold for multiple scattering paths. inliers C represents the cost function corresponding to line-of-sight distance and single scattering path. outliers This represents the cost function corresponding to multiple scattering paths;

[0104] The combination of minimum cost functions is labeled as The path index of the combination corresponding to the minimum cost function is denoted as Based on this, it was passed again The user's location, clock offset, and user orientation are estimated to obtain the state of the user terminal (UE).

[0105] Step 4. Based on the robustly jointly estimated UE state values ​​described above, iterative optimization is performed using the Gauss-Newton algorithm to obtain more refined user-end UE state estimates, as follows:

[0106]

[0107] Θ = argmin Θ E(Θ)

[0108] Among them, the parameter values ​​of the LosS path For the parameter values ​​of the k-th NLoS path Therefore, it satisfies E(Θ) represents the fitting error during the iteration process of the Gauss-Newton algorithm, z k Let Θ represent the measurement value of the k-th path, and let Θ represent the parameter value to be iteratively optimized, defined as... in express All paths except Loss. The parameter value Θ to be optimized is obtained by minimizing E(Θ).

[0109] Step 5: Define the interior points to be identified. There are M paths in total. Since each of these M paths between the BS and the UE could be a Loss-of-Stake (LoS) path, we construct M+1 candidate models for the LoS paths, where M represents the number of LoS cases and 1 represents NLoS cases. For each of the M LoS cases, there is one path that uses h... LoS (Θ LoS ) to parameterize the measured values, using The fitting process yielded M+1 sets of error values ​​E(Θ), based on the Akaike Information Criterion (AIC) strategy:

[0110]

[0111] AIC mΘ represents the AIC value of the m-th model fit. (m) E represents the parameter value estimated by the m-th model. m The fitting error of the m-th model, |Θ (m) | represents the dimension of the m-th model. Let m represent the number of interior points, where m ∈ M.

[0112] For M+1 AIC values, there exists a minimum AIC value, and the candidate model corresponding to the minimum AIC value can clearly identify the LoS and NLoS-1 paths.

Claims

1. A robust snapshot radio frequency SLAM method independent of multipath amplitude, used in a system formed by a base station and a user, characterized in that, Includes the following steps: S1. Using the base station location, user location, scattering point location, and the departure angle AoD and arrival angle AoA information of the signal propagation between the base station, user and scattering point, construct an amplitude-independent single snapshot geometric measurement model. In the single snapshot geometric measurement model, geometric mapping is used to form a model based only on the line-of-sight path LoS and the single non-line-of-sight path NLoS-1. S2. Based on the single snapshot geometric measurement model, the user position, system clock deviation and user orientation information are jointly estimated and then subjected to primary joint estimation and robust joint estimation to realize snapshot radio frequency SLAM. S2.1 Initial Joint Estimation: Using a single snapshot geometric measurement model, based solely on the channel parameter information of the line-of-sight path LoS and the single non-line-of-sight path NLoS-1, an initial joint estimation of user location, system clock offset, and user direction is performed. S2.2 Robust Joint Estimation: Based on the aforementioned primary joint estimation method, measurement information including multiple scattering paths is further introduced; using the RANSAC algorithm, the effective subset of measurements contributed by line-of-sight and single scattering paths is robustly selected from the complete measurement set containing line-of-sight LoS, single scattering NLoS-1, and multiple scattering paths. Based on the selected valid measurement subset, a final robust joint estimation of user position, system clock skew, and user orientation is performed.

2. The multipath amplitude-independent robust snapshot RF SLAM method according to claim 1, characterized in that: The process of constructing an amplitude-independent single-snapshot geometric measurement model is as follows: On the XYZ spatial coordinate axis, the single scattering volume point p is formed by the intersection of the i-th single scattering path. i With the coordinates of the point p as the center, and the single scattering point p as the center. i Distance from base station coordinates p BS and user coordinates p UE Construct two concentric circles with radius p. i To user coordinates p UE The distance and p i to base station coordinates p BS The distance, line segment (p) i p UE The radius of ) is smaller than that of line segment (p) i p BS The radius of p BS and p UE Located on two circles respectively, p BS to p i The extensions of the line segments intersect p UE The focus of the circle is point E, p UE to p i The extensions of the line segments intersect p BS Let point A be the focus of the circle, and let p... BS to p i The unit vector is denoted as u i That is, the direction vector of AOD, p i to p UE The unit vector is denoted as v i That is, the direction vector of AOA, and the vector containing the line segment (A, E) is denoted as vector m. i To simplify the calculation, the result of the cross product of vectors in three-dimensional space is projected onto the plane of the XY axis, that is, the Z-axis component is set to 0 for calculation. p BS With p i and p UE The included angle φ formed by the lines connecting them i -φ LoS With point A and p i The angle φ formed by the lines connecting point E and point E i ' are equal, where φ LoS This represents the departure angle of the line-of-sight path, therefore vector m i and v i The cross product of vector u is equal to that of vector u. i The cross product of t and t, where t = p UE -p BS The geometric relationship between the coordinate difference between the base station and the user satisfies: d i v i ×m i =d i u i ×t Where d i =c(τ) i -b UE ) represents the propagation distance of the i-th path, τ i Let c represent the propagation time of the i-th single path, and c represent the speed of light. Based on the geometric relationship of concentric circles, m i Rewritten as: m i =d i u i -(t+d i v i ) Where, m i This represents the vector containing the line segment AE from point A to point E; d i v i ×m i =d i u i The ×t rearrangement constructs an amplitude-independent single-snapshot geometric measurement model under noise-free conditions as follows: d i (u i ×v i )=t×(u i +v i )。 3. The multipath amplitude-independent robust snapshot RF SLAM method according to claim 2, characterized in that: The primary joint estimation method in step S2.1 is as follows: A single snapshot geometric measurement model based on noise-free conditions with independent amplitude. i (u i ×v i )=t×(u i +v i Introducing residual r under noisy conditions i (x UE ): in Using the least squares (LS) method to obtain information about x UE The optimal estimate minimizes the least squares criterion of the residuals; when considering the measurements Z = [z1, z2, ..., z] of all base station to user propagation paths. n When using the likelihood function p(z) of the channel parameters... i ∣Θ i )=N(z i |h i (Θ i ),Σ i ), by analyzing the mean h i (Θ i Apply noise Σ i Multiple sets of measurements Z can be generated; for all paths, the following values ​​about x are constructed. UE Cost function: The optimal estimate is: Where I = [1,2,...,n] represents the path index value of the measurement Z for all n paths.

4. The multipath amplitude-independent robust snapshot RF SLAM method according to claim 3, characterized in that: The robust joint estimation method in step S2.2 is as follows: Suppose the index of the measurement value Z for n paths is... By finding a Z subset that does not contain information about multiple scattering paths, and then using the Z subset to estimate x UE ; Use the RANSAC algorithm to randomly select N from the measured values ​​Z min The total number of combinations L, which combines paths, is expressed as: in, The index of the path represents the index value of the path in all combinations; Nmin represents the minimum number of paths required to solve for x; Let l be a combination within L, and let the path index corresponding to the l-th combination be denoted as . For the l-th combination A path, defined by the formula: Solve for the estimated value of the l-th combination; Due to the path index of the measured value Z Including line-of-sight, single scattering path, and multiple scattering path, using the formula right The division is as follows: in Indicates in The Loss of Scattering (LoS) paths and single scattering paths, or only single scattering paths, are divided into interior points, J. i (x)≥T ε The path whose cost function is greater than the threshold is a multiple scattering path. The path is denoted as the multiple scattering path information, called the outgoing point, and denoted as the multiple scattering path. express Within the line-of-sight distance and single scattering path; based on Recalculation passed The cost function C of the l-th combination is obtained. l : Where T ε C represents the detection threshold for multiple scattering paths. inliers C represents the cost function corresponding to line-of-sight distance and single scattering path. outliers This represents the cost function corresponding to multiple scattering paths; The combination of minimum cost functions is labeled as The path index of the combination corresponding to the minimum cost function is denoted as Based on this, it was passed again The user's location, clock offset, and orientation are estimated to obtain the user's UE state.

5. The multipath amplitude-independent robust snapshot RF SLAM method according to claim 4, characterized in that: Based on the robustly jointly estimated UE state values ​​described above, a more refined estimate of the user UE state is obtained through iterative optimization using the Gauss-Newton algorithm, as follows: Θ=argmin Θ E(I) Where E(Θ) represents the fitting error during the iteration process of the Gauss-Newton algorithm, z k Let Θ represent the measurement value of the k-th path, and let Θ represent the parameter value to be iteratively optimized, defined as... in express All paths except Loss; the parameter value Θ to be optimized is obtained by minimizing E(Θ); LoS path parameter values For the parameter values ​​of the k-th NLoS path Therefore, it satisfies h k The parameterization of the measurement value of the k-th path is represented as follows: If the k-th path is a Loss-free path: Where p BS and p UE These are the location coordinates of the base station and the user, respectively, α BS α UE and b UE These are the clock offsets in the base station direction, the user direction, and the system direction, respectively. and Let c be the transpose of the matrix, and c be the speed of light. If the k-th path is an NLoS path: Where p k The coordinates of the k-th path scattering point are given; the other parameters are the same as those for the line-of-sight path.

6. The multipath amplitude-independent robust snapshot RF SLAM method according to claim 5, characterized in that: The process of identifying LoS and NLoS-1 paths is as follows: Let the interior points be identified. There are M paths in total. Since each of these M paths between the base station (BS) and the user (UE) could potentially be a Loss-of-Stake (LoS) path, we construct M+1 candidate models for each LoS path, where M represents the number of LoS cases and 1 represents NLoS cases. For each of the M LoS cases, there is one path that uses h... LoS (Θ LoS ) to parameterize the measured values, using The fitting process yielded M+1 sets of error values ​​E(Θ), based on the Akaike Information Criterion (AIC) strategy: AIC m Θ represents the AIC value of the m-th model fit. (m) E represents the parameter value estimated by the m-th model. m The fitting error of the m-th model, |Θ (m) | represents the dimension of the m-th model. Let m represent the number of interior points, where m ∈ M; Find the smallest AIC value among M+1 AIC values. The candidate model corresponding to the smallest AIC value clearly identifies the LosS and NLoS-1 paths.

7. A computer device, characterized in that, It includes a processor and a memory, the processor being electrically connected to the memory, the memory being used to store instructions and data, and the processor being used to execute the multipath amplitude-independent robust snapshot RF SLAM method according to any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed by the multipath amplitude-independent robust snapshot RF SLAM method according to any one of claims 1-7.

Images