NLOS signal error identification method and system based on confidence circle algorithm

Abstract

The invention discloses an NLOS signal error identification method and system based on a confidence circle algorithm, and relates to the technical field of ranging networking, the proportion of the number of estimated nodes in a confidence circle to the total number of estimated nodes is used as an estimated node density, then all other estimated nodes are traversed, all estimated node densities are calculated, and an NLOS signal error identification result is obtained. Selecting the maximum estimated node density value from all the estimated node density values, taking the maximum estimated node density value as a final density value, judging whether a distance measurement value is interfered by an NLOS error or not according to the obtained estimated node density when the to-be-measured node and a new node initiate distance measurement, and if the estimated node density is less than a certain threshold value, judging whether the distance measurement value is interfered by the NLOS error or not; and if not, determining that the estimated node position outside the confidence circle has an NLOS error, reducing the weight of the identified NLOS signal according to the final density value, and processing the NLOS signal by using one-dimensional Kalman filtering. The identification method is provided.

Description

Technical Field

[0001] This invention relates to the field of ranging networking technology, specifically to a method and system for identifying NLOS signal errors based on the confidence circle algorithm. Background Technology

[0002] In satellite-denied environments, such as jungles, indoor spaces, and mines, existing satellite navigation systems can no longer provide users with accurate location information. Ultra-wideband (UWB) signals, due to their high positioning accuracy and strong multipath resistance, are often used to provide location services in these scenarios. Common UWB positioning algorithms, such as Time of Arrival (TOA) positioning, rely on accurate ranging results. However, UWB signals are highly susceptible to interference from obstructions, generating Non-Line of Sight (NLOS) signals, which in turn introduce NLOS errors into the ranging results. Therefore, researching methods for identifying and suppressing NLOS errors is crucial for improving the accuracy and reliability of UWB positioning systems. In recent years, research on error identification and suppression of ultra-wideband NLOS signals has attracted increasing attention. Machine learning, due to its advantages in nonlinear fitting, is often applied to error identification of ultra-wideband NLOS signals. With the rapid development of computer technology, deep learning theory has been widely used in target recognition and classification. Convolutional Neural Network (CNN) is an effective method for target recognition and classification, but due to its poor real-time performance and dependence on prior environmental information, CNN cannot build a good model for error identification of ultra-wideband NLOS signals. To address this deficiency, a Transformer model has been developed that effectively captures long-range multipath correlations in ultra-wideband signals by introducing an efficient channel attention mechanism. By assigning different weights to data at different times, the algorithm's adaptability to various environments is enhanced. However, in terms of error suppression for ultra-wideband NLOS signals, wavelet filtering can preserve the localized time-frequency characteristics of non-stationary signals, separating error signals from effective signals and selectively suppressing error components. This maximizes the preservation of effective signal details in the low-frequency subband, allowing for NLOS signal error suppression through signal reconstruction. However, the effectiveness of wavelet filtering largely depends on the choice of wavelet basis. When the wavelet basis does not match the signal characteristics, it will filter out effective components in the signal, thus reducing the filtering effect.

[0003] Based on this, the present invention proposes a method and system for identifying NLOS signal errors based on the confidence circle algorithm. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for identifying NLOS signal errors based on the confidence circle algorithm, so as to solve the problems in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an NLOS signal error identification method based on the confidence circle algorithm, the identification method comprising the following steps: Step 1: Select a node as the node to be tested. Divide the nodes around the node to be tested into a subgroup of at least three, for a total of N groups. Calculate the estimated node coordinates of the node to be tested in each subgroup using the least squares method, and calculate the distance between the node to be tested and its adjacent nodes. Step 2: Using the actual location of the estimated node as the center and the average distance deviation between the estimated and actual locations of all estimated nodes as the radius, obtain a confidence circle. Count the number of estimated nodes within the confidence circle and use the proportion of estimated nodes within the confidence circle to the total number of estimated nodes as the estimated node density. Then, iterate through all other estimated nodes, calculate the density of all estimated nodes, and select the largest estimated node density value from all estimated node density values. , and as the final density value; Step 3: When the node to be measured initiates distance measurement with a new node, the distance measurement value is determined based on the estimated node density. If the estimated node density is less than a certain threshold, it is considered that the estimated node position outside the confidence circle has NLOS error. The weight of the identified NLOS signal is reduced based on the final density value, and a one-dimensional Kalman filter is used for processing.

[0006] In a preferred embodiment, step three, reducing the NLOS signal weights, includes the following steps: Set the LOS ranging value weight to The weights of the NLOS ranging values ​​are set to ,in To provide a real-time estimated node density for the current node under test. This represents the maximum estimated node density obtained by traversing all nodes.

[0007] In a preferred embodiment, in step three, it is determined whether the distance measurement value is interfered with by NLOS error. The mathematical model is expressed as follows: ,in, The radius of the confidence circle; This represents the confidence coefficient, and If the output is If the distance measurement is affected by the NLOS signal, then the output will be... This indicates that it was not affected by NLOS error.

[0008] In a preferred embodiment, the node to be tested is any one of the drones in the drone swarm.

[0009] This application also provides an ultra-wideband NLOS signal error identification system based on the confidence circle algorithm, including an ultra-wideband signal generation module, an ultra-wideband positioning module, and a data processing module; The ultra-wideband signal generation module is used to generate or receive narrow pulse signals with a time width on the order of nanoseconds; The ultra-wideband positioning module calculates the distance between modules by measuring the transmission and reception time of the ultra-wideband signal, and then calculates the location information based on the distance information. The data processing module is used to receive distance and position data output by the ultra-wideband positioning module, construct a confidence circle to determine whether there is NLOS error in the signal, and after identifying the NLOS signal, suppress the NLOS signal through one-dimensional Kalman filtering. The data processing module performs the steps of the identification method according to any one of claims 1-3 to complete the identification and suppression of NLOS signal errors.

[0010] In a preferred embodiment, the ultra-wideband signal generation module includes a pulse generator, a signal amplification circuit, and a control unit, wherein the control unit evaluates the quality of the received signal based on the parameters of the received signal.

[0011] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0013] Figure 1 This is a schematic diagram of the positioning principle of the present invention; (a) TOA positioning principle diagram; (b) NLOS error diagram; Figure 2 A graph showing the relationship between confidence coefficient and recognition rate; Figure 3 Indoor and outdoor fixed-point NLOS signal identification diagram; (a) Outdoor distance measurement results; (b) Indoor distance measurement results diagram; Figure 4 Filtered image of ranging results. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example: Please refer to Figures 1-4 As shown, this embodiment provides a method for identifying NLOS signal errors based on the confidence circle algorithm: The identification method specifically includes the following steps: Step 1: Select a node as the node to be tested, and divide its surrounding nodes into subgroups of three, for a total of N groups. Calculate the estimated coordinates of the node to be tested in each subgroup using the least squares method, and calculate the distance between the node to be tested and its adjacent nodes.

[0016] Using the observed distances from each reference node to the node to be measured as known quantities and the coordinates of the node to be measured as unknown quantities, a distance equation is established (i.e., the geometric distance from each reference node to the node to be measured is equal to the measured value). Since this equation is nonlinear, it is usually necessary to linearize it near an initial estimated location of the node to be measured (e.g., the centroid of the reference node's coordinates). This involves performing a first-order Taylor expansion of the nonlinear equation at the initial point and ignoring higher-order terms, thus transforming the problem into a linear least squares problem. Next, a system of linear equations incorporating the observation residuals is constructed, and the coordinate corrections are solved using matrix operations (such as the least squares formula). Through multiple iterations, the true coordinates of the node to be measured are gradually approximated.

[0017] Step 2: Using the actual location of the estimated node as the center and the average distance deviation between the estimated locations and the actual locations of all estimated points of the node to be measured as the radius, obtain a confidence circle. Count the number of estimated points within the confidence circle and use the proportion of estimated points within the confidence circle to the total number of estimated points as the estimated point density. Then, iterate through all other nodes, calculate the density of all estimated points, and select the highest estimated point density value from all the estimated point density values. And use this density value as the final density value.

[0018] The specific calculation method for the confidence circle radius R includes: The radius is based on the Euclidean distance deviations between the estimated positions of all test nodes obtained through the least squares method and a selected reference position (e.g., true coordinates, high-precision estimated coordinates, or the center position of all estimated points). The average of these deviations is calculated and used as the radius of the confidence circle. The geometric center of all estimated points, or a known high-precision reference node coordinate, is selected as the benchmark.

[0019] The confidence coefficient is a crucial parameter used in this application to measure the relationship between the distribution of estimated points within the confidence circle and the likelihood of NLOS error. Essentially, it is an empirical or experimentally calibrated constant or adjustable parameter, typically ranging from 0 to 1. In this application, the confidence coefficient and the radius of the confidence circle jointly participate in the judgment model, primarily used to adjust the sensitivity to NLOS error: when the confidence coefficient is small, the system is more sensitive to sparse estimated points within the confidence circle, making it easier to determine the presence of NLOS error; conversely, when the confidence coefficient is large, the system is more lenient in its requirements regarding the distribution of estimated points, only identifying NLOS interference when the estimated points are extremely sparse.

[0020] In practical applications, this confidence coefficient is not arbitrarily set, but rather an empirical parameter calibrated through extensive simulation experiments or measured data. Specifically, a series of test scenarios with known NLOS interference and pure LOS (line-of-sight) environments can be constructed to simulate ranging data under different degrees of occlusion, different node distribution densities, and different ranging error levels. The estimated point density within the corresponding confidence circle can be calculated, and the system's ability to correctly identify NLOS signals (such as recognition rate, false positive rate, etc.) under different confidence coefficient values ​​can be statistically analyzed. Based on these simulation results, a curve showing the relationship between the confidence coefficient and the NLOS signal recognition rate can be plotted (as shown in the attached figure of this application specification). Figure 2 As shown in the figure, a confidence coefficient that can both ensure a high NLOS recognition rate and control the false positive rate is selected as the recommended value.

[0021] Experiments show that the typical range of confidence coefficient values ​​is between 0.2 and 0.6, and the specific value can be fine-tuned according to the characteristics of the actual deployment environment. For example, in indoor environments, where node distribution is relatively dense and the types of obstructions are relatively fixed, the confidence coefficient can be set to a smaller value (e.g., around 0.3) to improve the detection sensitivity to local NLOS interference. In outdoor open environments or scenarios with many dynamic obstructions (e.g., trees, crowds), to avoid misjudging normal fluctuations as NLOS errors, the confidence coefficient can be set to a relatively larger value (e.g., 0.5 or higher). In addition, if the system has adaptive adjustment capabilities, the confidence coefficient value can also be dynamically adjusted based on the historical ranging error statistics or real-time estimated point distribution trends under the current environment to further improve the robustness of NLOS error identification.

[0022] The estimated point density threshold is another key parameter used in this application to determine whether there is NLOS error in the current ranging. It is defined as follows: when a node to be measured is ranging with other nodes, if the proportion of the number of estimated points within the confidence circle to the total number of estimated points (i.e., the estimated point density) is lower than this threshold, then it is considered that there is significant NLOS error interference in this ranging, and the corresponding ranging value needs to be downweighted or further filtered. This threshold is essentially a dividing line between "normal ranging" and "ranging that may be affected by NLOS interference," and its reasonable setting is crucial to avoiding misjudgment and missed judgment.

[0023] The original specification mentioned that non-line-of-sight signals are identified by setting an estimated point density threshold. When the estimated point density is lower than the threshold, it is determined that a non-line-of-sight signal exists in the ranging measurement. However, it did not specify whether the threshold is a fixed value or a dynamic value, nor did it explain the specific basis for its value or experimental verification. To enable those skilled in the art to actually set this parameter, the following supplementary explanation is provided: In practical applications, the estimated point density threshold can be either a fixed value determined based on a large amount of experimental data or a relative value dynamically calculated based on the current environmental characteristics. If a fixed threshold strategy is adopted, an empirical value that can effectively distinguish between normal ranging and NLOS interference can generally be selected based on the estimated point density distribution in typical test scenarios. For example, in most simulation experiments and measured data, when the estimated point density within the confidence circle is less than 30% to 40% of the total estimated point density, it often means that the estimated point distribution in that area is relatively sparse, and the possibility of NLOS error increases significantly. Therefore, a typical fixed value for the estimated point density threshold can be set to 30%, 35%, or 40%. This type of fixed threshold is suitable for scenarios with relatively stable environments and strong regularity in node distribution, such as indoor fixed base station deployment or collaborative positioning of drone swarms in known open areas.

[0024] If the deployment environment is complex and frequently changing (e.g., the drone constantly changes its relative position during flight, urban canyon environments, or the presence of moving obstructions), a dynamic threshold strategy is recommended. This involves calculating a dynamic reference density (e.g., a certain percentage of the average, median, or maximum density of all estimated points) based on the overall distribution of all estimated points at the current moment, and setting the estimated point density threshold to a certain percentage of this reference density (e.g., 50% or 60%). In this way, the system can automatically adjust its judgment criteria according to the actual distribution density of estimated points in the environment, thereby improving the adaptability and accuracy of NLOS error identification.

[0025] Regardless of whether a fixed or dynamic threshold is used, its final value should be determined through experimental verification. Specific experimental methods may include: in a test environment with a known mix of LOS and NLOS, statistically analyzing the distribution of ranging errors and recognition results under different estimated point densities, identifying the optimal threshold range that balances recognition rate and false positive rate, and incorporating typical values ​​within this range as recommended thresholds into the system configuration or manual. Furthermore, if the system possesses adaptive learning capabilities, it can automatically optimize and update the estimated point density threshold based on historical ranging data and error statistics to adapt to environmental changes.

[0026] Step 3: When the node under test initiates distance measurement with a new node, based on the obtained estimated point density, it is determined whether the distance measurement value is interfered with by NLOS error. If the estimated point density is less than a certain threshold, it is considered that the estimated point position outside the confidence circle has NLOS error. The weight of the identified NLOS signal is reduced and processed using one-dimensional Kalman filtering.

[0027] In this application, the one-dimensional Kalman filter primarily processes the ranging value to suppress NLOS errors. Therefore, the state variable is typically defined as a single ranging value, i.e., the ranging observation between the target node and the reference node at the current moment. In most application scenarios, this ranging value is considered the only state variable to be estimated, without introducing other state variables such as the rate of change of ranging, velocity, or acceleration. This simplified one-dimensional state definition effectively reduces the computational complexity of the filter while meeting the real-time and resource consumption requirements of ultra-wideband positioning systems.

[0028] Of course, if there is a higher requirement for tracking the dynamic changes of the ranging value in practical applications (such as high-speed movement of the UAV or significant relative motion between nodes), the ranging change rate (i.e., the first derivative of the ranging value) can be used as an auxiliary state variable to construct a two-dimensional state vector. However, in the typical implementation described in this specification, to ensure the universality and ease of implementation of the solution, the state variables by default only contain the ranging value itself, that is, a one-dimensional state.

[0029] The core of one-dimensional Kalman filtering lies in describing the evolution of the system state (i.e., the measured value) over time through a state transition equation, and correlating the actual measured value with the system state through an observation equation. Although specific mathematical formulas are not given in this application, its basic principle can be described as follows: State transition equations assume that the ranging value changes slowly over a short period of time, meaning the change is close to zero or fluctuates within a certain range. Therefore, the state transition equations typically assume that the current ranging value is mainly derived from the ranging value of the previous time step using a simple dynamic model (such as a constant velocity model or a constant value model). Under a constant value model, the ranging value can be considered to remain constant in the short term; that is, the ranging value at the next time step equals the current ranging value plus a small random disturbance (described by process noise). This modeling approach is simple and effective, and applicable to most static or quasi-static ranging environments.

[0030] Observation equations: Observation equations are used to relate the actual ranging observations (i.e., the raw ranging data obtained through ultra-wideband signals, or the ranging values ​​adjusted by NLOS signal weights) to the system state (i.e., the ranging values ​​estimated by the filter). In this application, the observed value is the ranging measurement value obtained at the current moment, and the observation equations describe how this measurement value is obtained by adding a certain amount of observation noise to the system state (ranging value). In short, the observation equations reflect the basic idea that "the value we measure through the sensor is the true value plus noise."

[0031] The aforementioned state transition equation and observation equation together form the basis of the two core steps of Kalman filtering: prediction and update. This enables the filter to recursively estimate the most likely distance value at each time step based on prior information (prediction) and new observations (update).

[0032] III. Statistical Characteristics of Process Noise and Observation Noise In one-dimensional Kalman filtering, process noise and observation noise are two key statistics, describing the uncertainties in the system's state evolution and the random errors in sensor measurements, respectively. Their statistical characteristics directly affect the filter's performance, especially its smoothness and response speed.

[0033] Process noise (Q): Process noise characterizes the random fluctuations that may occur in the ranging value during its evolution over time, such as errors caused by small node movements, environmental perturbations, or model simplification. In this application, process noise is typically assumed to follow a Gaussian distribution with zero mean, and its variance is denoted as Q. The typical range of Q needs to be determined based on the actual application scenario: in environments where nodes are basically stationary or move slowly, Q can take a smaller value (e.g., on the order of 1e-4 to 1e-2) to reflect the relative stability of the ranging value; while in scenarios where nodes move quickly or the environment changes significantly, Q can be appropriately increased (e.g., on the order of 1e-2 to 1e-1) to allow the filter to track changes in the ranging value more quickly.

[0034] Observation noise (R): Observation noise describes the measurement uncertainty caused by factors such as signal reflection, multipath effects, and hardware errors during actual ranging. This noise is also modeled as a zero-mean Gaussian distribution, with its variance denoted as R. The value of R typically depends on the hardware performance of the ultra-wideband sensor, the ranging environment (e.g., indoor / outdoor, obstruction), and the accuracy of the signal processing algorithm. In practical settings, R can be obtained based on the variance statistics of the sensor's technical parameter manual, calibration experiments, or historical ranging data. The typical range may be between 1e-2 and 1, and the specific value needs to be determined through experiments or simulations.

[0035] The proper setting of Q and R is crucial for the tracking capability and stability of the filter. It usually needs to be optimized through experiments or simulations to suppress NLOS error while avoiding over-smoothing or tracking lag.

[0036] One-dimensional Kalman filtering requires initialization before its first run, which involves setting the initial state estimate and the initial error covariance matrix. These initial values ​​serve as the starting point for the filter's recursive calculations and have a certain impact on the filter's convergence speed and initial accuracy.

[0037] Initial state estimate: This can typically be set to the original ranging value obtained from the first ranging measurement, or an initial ranging value estimated based on the node's initial position information, reference node coordinates, and communication delay. In the absence of prior information, the initial state estimate can also be set to zero or a reasonable default value, and then quickly converged to near the true value through the filter's prediction and update process.

[0038] Initial error covariance matrix (scalar in one-dimensional case): used to describe the uncertainty of the initial state estimate. If the initial ranging value comes from a high-precision positioning method (such as GPS+IMU fusion positioning), the initial error covariance can be set to a small value (e.g., on the order of 1e-2 to 1) to reflect the high reliability of the initial estimate; conversely, if the initial ranging value comes from an uncalibrated sensor or the first estimation, its uncertainty is high, and the initial error covariance can be set to a large value (e.g., on the order of 1 to 1e2) to give the filter more adjustment space.

[0039] After initialization, the filter will enter a regular prediction and update loop, continuously optimizing the estimation of the distance value based on the new observations at each time step.

[0040] V. Adaptive Mechanism of Adaptive Kalman Filtering (if adopted) In applications with strong dynamic changes or unstable ranging environments, this application can also employ adaptive Kalman filtering technology to improve the adaptability of one-dimensional Kalman filtering to time-varying noise and system uncertainties. The core of adaptive Kalman filtering lies in dynamically adjusting the values ​​of process noise variance Q and observation noise variance R based on real-time observation data, thereby enabling the filter to automatically optimize its performance according to the current environmental conditions.

[0041] The basic principle of adaptive mechanisms typically includes: by monitoring the magnitude and trend of the residuals (i.e., the difference between the observed and predicted values), it can be determined whether there are any abnormalities in the current observed noise or system state changes; if a significant increase in the residuals is detected, it indicates that the current observed noise may be increasing or the system state may be undergoing a sudden change. In this case, the adaptive algorithm will increase the observed noise variance R or the process noise variance Q accordingly to allow the filter to track changes more flexibly; conversely, if the residuals remain small and stable, it indicates that the current environment is relatively stable, and the values ​​of Q and R can be appropriately reduced to improve the smoothness and accuracy of the filtering.

[0042] like Figure 1 As shown in (a), in step one, the node to be tested needs to obtain the coordinates of its surrounding nodes. Assume the coordinates of the node to be tested are... The coordinates of the three reference nodes of a subgroup are as follows: , , Three nodes to node The positions are respectively Using the position of each node as the center and the distance from each node to the node to be measured as the radius, three circles will be obtained. According to the solution principle, the intersection of the three circles is the solution point. Location. The specific calculation process is as follows.

[0043] (1) Expanding the above formula, we get: (2) set up Then we have: (3) The above formula can be written in matrix form: (4) The above formula can then be further written as: .

[0044] Coordinates of the node to be measured That is, the obtained Using the least squares method, we can obtain: (5) like Figure 1 As shown in (b), in step two, when there is non-line-of-sight (LOS) between nodes, the distance measurement between nodes will be affected by LOS errors. In this study, this LOS interference is called "error force," and the point obtained by the node under test through the above calculation is called the estimated point. When three nodes measure the distance of the node under test, if one node has an "error force," the estimated point will be far away from the estimated point obtained without an "error force," and the number of outliers in the estimated point group of the node under test will increase. Conversely, in a line-of-sight environment, no node has an "error force," and the estimated points will exhibit a clustering phenomenon. Therefore, LOS and NLOS signals can be identified by calculating the proportion of the number of estimated points in a certain area around the node under test to the total number of estimated points. The proportion of the number of estimated points in a certain area to the total number of estimated points is called the estimated point density, and this area is called the confidence circle.

[0045] Based on the measurement data between the node under test and other nodes, a confidence circle is obtained, and the estimated point density is calculated. Non-line-of-sight (NLS) signals can be identified by setting a threshold for the estimated point density. When the estimated point density is below the threshold, it is determined that a NLS signal exists in that ranging measurement. Conversely, if the estimated point density is above the threshold, it is considered that no NLS signal exists in that ranging measurement.

[0046] Using the actual location of the estimated node as the center and the average distance deviation between the estimated locations and the actual locations of all estimated points of the node to be measured as the radius, a confidence circle is obtained, and its radius R is calculated. The number of estimated points within the confidence circle is counted, and the proportion of the number of estimated points within the confidence circle to the total number of estimated points is taken as the estimated point density. Then, iterate through all other nodes, calculate the density of all estimated points, and select the highest estimated point density value from all the estimated point density values. And use this density value as the final density value.

[0047] When the node under test initiates a distance measurement with a new node, it is determined whether the distance measurement value is interfered with by NLOS error based on the obtained estimated point density. If the estimated point density is less than a certain threshold, it is considered that the estimated point position outside the confidence circle has NLOS error, and there is NLOS signal in the wireless propagation process between it and the node under test.

[0048] The mathematical model for determining the presence of an NLOS signal can be expressed as: (6) in, The radius of the confidence circle; This represents the confidence coefficient, and If we assume If the result is true, the distance measurement is affected by the NLOS signal; otherwise, it is not affected by NLOS error. Therefore, for this algorithm, the radius of the confidence circle is different each time, and its value is uncertain. The value of should be a determinable parameter. By altering the propagation delay within the unit circle, the existence of non-line-of-sight signals can be simulated. The recognition rate of the NLOS signal was obtained by taking values ​​from 0.1 to 1. The specific simulation results are shown in the figure below. Figure 2 As shown.

[0049] In step three, the weight of the identified NLOS ranging values ​​in the positioning calculation is reduced: let the weight of the LOS ranging value be... The weight of the NLOS ranging value is ( The smaller, The lower the value, the smaller the impact of NLOS data.

[0050] For the weighted distance measurement sequence One-dimensional Kalman filtering is used to suppress NLOS error. The filtering model is as follows: Equations of state: (Assuming the distance measurement value changes at a constant speed,) The process noise follows the pattern N(0,Q), where Q is the variance of the process noise. Observation equation: (z) k The weighted distance measurement values ​​are the observed values. The observation noise follows the pattern N(0,R), where R is the variance of the observation noise. Iterative calculation of filter gain State estimation and estimation error covariance High-precision ranging value after output filtering .like Figure 3 As shown in (a) and (b), the filtering results of the filtering algorithm in indoor and outdoor experiments are shown respectively. Figure 4 The figure shown represents the results of the filtering algorithm in a simulation environment.

[0051] An ultrawideband NLOS signal error identification and suppression system based on the confidence circle algorithm: The prediction system includes: an ultra-wideband signal generation module, an ultra-wideband positioning module, and a data processing module; An ultra-wideband signal generation module is used to generate or receive narrow pulse signals with a time width on the nanosecond level. Its core components include a pulse generator, a signal amplification circuit, and a control unit. The quality of the received signal can be evaluated based on the parameters of the received signal. The ultra-wideband positioning module accurately calculates the distance between modules by measuring the time of transmission and reception of ultra-wideband signals, and further calculates the location information using the distance information. The data processing module is used to construct a confidence circle from the measured data input to determine whether the signal exhibits NLOS. Based on the identification of NLOS signals, a one-dimensional Kalman filter is used to suppress the NLOS signals.

[0052] A drone swarm, wherein the drone swarm is equipped with the aforementioned NLOS signal error identification and suppression system.

[0053] Step 1: Drone Cluster Node Grouping and Initial Positioning Estimation. Select any one drone in the cluster as the node to be tested. Using the broadcast function of the ultra-wideband signal generator module, discover neighboring drones within communication range and use them as reference nodes. Each drone obtains its own high-precision absolute coordinates through GPS+IMU fusion positioning and sends these coordinates to the node to be tested via the ultra-wideband signal generator module. For neighboring nodes, divide them into subgroups of three, and calculate the estimated coordinates of each subgroup using the least squares method.

[0054] Step 2: Confidence Circle Construction and Estimated Point Density Calculation. Based on the estimated point coordinates calculated in Step 1, construct the confidence circle, calculate the estimated point density, iterate through all drones in the cluster, calculate the estimated point density for drones without drones, and select the highest density as the final reference density for the cluster.

[0055] Step 3: NLOS Signal Identification and Suppression. When the node under test initiates ranging with a new reference node (or a moved neighbor), the estimated point density is calculated and compared with a preset threshold. If the density is less than the threshold, the node is determined to have received NLOS interference; otherwise, no NLOS interference is determined. The weight of the identified NLOS ranging values ​​in the positioning calculation is reduced. For the adjusted ranging value sequence, a one-dimensional Kalman filter is used to calculate the filtered high-precision ranging value.

[0056] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0057] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for NLOS signal error identification based on a confidence circle algorithm, characterized in that: The identification method comprises the following steps: Step one: select a node as a to-be-tested node, and divide the nodes around the to-be-tested node into N groups each with at least three nodes; calculate the estimated node coordinates of the to-be-tested node in each group according to the least square method, and calculate the distance between the to-be-tested node and the adjacent nodes; Step two: take the actual position of the estimated node as the center, and the average of the ranging deviation between all estimated node positions and the actual position as the radius, to obtain a confidence circle, count the number of estimated nodes within the confidence circle, and take the proportion of the number of estimated nodes within the confidence circle in the total number of estimated nodes as the estimated node density Then traverse all other estimated nodes, calculate the density of all estimated nodes, and select the maximum estimated node density value from all estimated node density values And take it as the final density value Step three: when the to-be-tested node initiates distance measurement with a new node, determine whether the distance measurement value is disturbed by NLOS error according to the obtained estimated node density; if the estimated node density is less than a certain threshold, it is considered that the estimated node position outside the confidence circle has NLOS error; reduce the weight of the identified NLOS signal according to the final density value, and process it using one-dimensional Kalman filtering.

2. The confidence circle algorithm based NLOS signal error identification method of claim 1, wherein: In step three, reducing the weight of the NLOS signal comprises the following steps: The LOS ranging value weight is set as , the NLOS ranging value weight is set as , wherein is the real-time estimated node density of the current to-be-measured node, is the maximum estimated node density obtained by traversing all nodes.

3. The NLOS signal error identification method based on the confidence circle algorithm of claim 2, characterized in that: In step three, it is determined whether the distance measurement is interfered by NLOS error, and the mathematical model is expressed as: , wherein, represents the radius of the confidence circle; represents the confidence coefficient, and , if the output is , the distance measurement is affected by NLOS signal, and if the output is , it is indicated that it is not interfered by NLOS error.

4. The confidence circle algorithm based NLOS signal error identification method of claim 1, wherein: The to-be-tested node is any one of the unmanned aerial vehicle cluster.

5. An ultra-wideband NLOS signal error identification system based on a confidence circle algorithm, characterized by: It comprises an ultra-wideband signal generation module, an ultra-wideband positioning module and a data processing module. The ultra-wideband signal generation module is used to generate or receive narrow pulse signals with a time width of nanoseconds; The ultra-wideband positioning module calculates the distance between modules through the sending and receiving time of the ultra-wideband signal, and calculates the position information according to the distance information; The data processing module is used to receive the distance and position data output by the ultra-wideband positioning module, construct a confidence circle to determine whether there is NLOS error, and after identifying the NLOS signal, suppress the NLOS signal through one-dimensional Kalman filtering; The data processing module executes the steps of the identification method in any one of claims 1-3 to complete the identification and suppression of NLOS signal error.

6. The confidence circle algorithm based ultra-wideband NLOS signal error identification system of claim 4, wherein: The ultra-wideband signal generation module comprises a pulse generator, a signal amplification circuit and a control unit, and the control unit evaluates the quality of the received signal according to the parameters of the received signal.

Images