Ultra-wideband positioning dimension adaptive switching method and system based on signal multipath feature recognition

Abstract

The application discloses a kind of based on signal multipath feature recognition ultra-wideband positioning dimension adaptive switching method and system, belong to ultra-wideband positioning technical field, this method includes: by extracting the multipath characteristic parameter in channel impulse response, including first path amplitude, multipath time delay spread, multipath component number and first path power ratio etc., input pre-trained scene classification model, identify out three kinds of scenarios of line of sight, non line of sight and dense multipath;According to scene adaptive switching to one-dimensional, two-dimensional or three-dimensional positioning mode, wherein dense multipath scene switches to three dimensions to utilize multi-base station redundancy to suppress error, line of sight scene switches to low dimension to save resources.The application only utilizes signal itself to realize pure drive adaptation, does not rely on additional sensor, and supports smooth switching and closed-loop feedback, can reduce system power consumption while ensuring positioning accuracy, suitable for high-precision positioning in indoor complex environment.

Description

Technical Field

[0001] This invention relates to the field of broadband hyperlocation technology, specifically to a method and system for adaptive switching of ultra-wideband positioning dimensions based on signal multipath feature recognition. Background Technology

[0002] The basic principle of ultra-wideband (UWB) positioning systems is to measure the signal propagation time between a positioning tag and multiple known-location base stations, calculate the distance between the tag and each base station, and then use multiple distance measurements to calculate the position coordinates of the target using a geometric intersection method. Depending on the positioning dimension, UWB positioning can be divided into one-dimensional positioning (one-dimensional coordinate calculation along a straight line or track), two-dimensional positioning (two-dimensional coordinate calculation in a plane), and three-dimensional positioning (three-dimensional coordinate calculation in space). Different positioning dimensions require different numbers of base stations, and there are significant differences in positioning accuracy and system resource consumption.

[0003] However, ultra-wideband signals are inevitably affected by multipath effects and non-line-of-sight (NLOS) propagation during actual transmission. In indoor environments, signals are reflected, refracted, and scattered by obstacles such as walls, furniture, and people, resulting in multiple propagation paths. In line-of-sight (LOS) environments, the direct path signal dominates, leading to higher positioning accuracy. In non-line-of-sight (NLOS) environments, signal reflection and refraction through obstacles cause time delays, resulting in signal clutter, inaccurate measurements, and significantly increased positioning errors. In dense multipath environments, numerous reflected paths overlap, potentially overwhelming or severely attenuating the initial path signal, posing even greater challenges to positioning calculations.

[0004] To address the aforementioned issues, several solutions already exist in the prior art. For example, Chinese invention patent CN110099352B discloses an intelligent switching method for positioning base stations required by a UWB positioning system in two-dimensional and three-dimensional positioning scenarios. This solution is based on the receiving angle of the positioning node. The positive and negative changes in the signal strength are used to determine which layer of base stations the location tag is located in, thereby achieving intelligent switching of the required location base station. However, this scheme relies on geometric angle parameters for switching, which essentially selects the location base station based on the relative spatial relationship between the tag and the base station, without considering the impact of the physical characteristics of the signal propagation environment on positioning accuracy.

[0005] Furthermore, existing technologies include line-of-sight (LOS) / non-line-of-sight (NLOS) identification based on Channel Impulse Response (CIR). For example, by analyzing parameters such as first path amplitude and multipath delay spread in the CIR, the current channel state can be determined as LOS or NLOS, thereby correcting the ranging error. However, the application scenarios of these schemes are usually limited to binary judgment of channel state (LOS / NLOS), and their output results are mainly used for ranging error correction, without involving adaptive switching of positioning dimensions.

[0006] Therefore, it is evident that existing ultra-wideband positioning systems have the following technical shortcomings: First, the positioning dimension is fixed or based solely on geometric information switching, making it impossible to adaptively adjust according to the physical characteristics of the actual signal propagation environment. In environments with good line-of-sight, the system may still use high-dimensional positioning, resulting in unnecessary waste of system resources and increased power consumption. In complex environments such as dense multipath propagation, if the system still uses low-dimensional positioning, it cannot fully utilize multi-base station information for redundant calculations to combat multipath errors.

[0007] Second, existing CIR-based scene recognition schemes only perform binary classification of LOS / NLOS, resulting in coarse recognition granularity. They cannot distinguish between scenes of different complexity, such as "ordinary non-line-of-sight" and "dense multipath", making it difficult to support refined positioning strategy decisions.

[0008] Third, existing solutions lack a smooth transition mechanism when switching positioning dimensions. Frequent or abrupt dimension switching can lead to jumps and discontinuities in positioning results, affecting user experience and positioning stability. Summary of the Invention

[0009] To address the technical shortcomings of existing ultra-wideband positioning systems, such as fixed positioning dimensions, unreasonable switching criteria, coarse scene recognition granularity, and lack of smooth switching mechanisms, this invention provides an ultra-wideband positioning dimension adaptive switching method and system based on signal multipath feature recognition.

[0010] In a first aspect, the present invention provides an ultra-wideband positioning dimension adaptive switching method based on signal multipath feature recognition, comprising the following steps: S1. Receive the ultra-wideband positioning signal transmitted by the positioning target through an ultra-wideband positioning receiver, obtain channel impulse response data, and extract multipath signal components from the channel impulse response data; S2. Perform time-domain and frequency-domain analysis on the extracted multipath signal components to extract multipath feature parameters that characterize the multipath propagation characteristics of the current positioning environment. The multipath feature parameters include at least the first path amplitude, the first path arrival time, the multipath delay spread, the number of multipath components, and the multipath power delay distribution. S3. Input the extracted multipath feature parameters into the pre-trained scene classification model, and the scene classification model identifies and classifies the current location scene, and outputs the scene classification result. The scene classification result includes at least line-of-sight scene, non-line-of-sight scene and dense multipath scene. S4. Based on the scene classification results, determine the target positioning dimension that matches the current positioning scene. The positioning dimension includes one-dimensional positioning mode, two-dimensional positioning mode and three-dimensional positioning mode. S5. Switch the ultra-wideband positioning system from the current positioning dimension to the target positioning dimension, and use the positioning calculation algorithm corresponding to the target positioning dimension to calculate the position of the positioning target; In step S5, when the scene classification result is a line-of-sight scene, the system switches to a one-dimensional positioning mode or a two-dimensional positioning mode; when the scene classification result is a non-line-of-sight scene, the system switches to a two-dimensional positioning mode; and when the scene classification result is a dense multipath scene, the system switches to a three-dimensional positioning mode.

[0011] Secondly, the present invention also provides an ultra-wideband positioning dimension adaptive switching system based on signal multipath feature recognition, comprising: The ultra-wideband signal receiving module is used to receive the ultra-wideband positioning signal transmitted by the positioning target through the ultra-wideband positioning receiver and obtain channel impulse response data; A multipath signal extraction module is used to extract multipath signal components from the channel impulse response data; The multipath feature extraction module is used to perform time-domain and frequency-domain analysis on the extracted multipath signal components and extract multipath feature parameters to characterize the multipath propagation characteristics of the current positioning environment. The multipath feature parameters include at least the first path amplitude, the first path arrival time, the multipath delay spread, the number of multipath components, and the multipath power delay distribution. The scene classification module is used to store pre-trained scene classification models, input the extracted multipath feature parameters into the scene classification models, and have the scene classification models identify and classify the current location scene and output scene classification results. The scene classification results include at least line-of-sight scenes, non-line-of-sight scenes and dense multipath scenes. The positioning dimension decision module is used to determine the target positioning dimension that matches the current positioning scene based on the scene classification results. The positioning dimension includes one-dimensional positioning mode, two-dimensional positioning mode and three-dimensional positioning mode. The positioning dimension switching module is used to switch the ultra-wideband positioning system from the current positioning dimension to the target positioning dimension; The positioning calculation module is used to calculate the position of the positioning target using a positioning calculation algorithm corresponding to the positioning dimension of the target. Specifically, the positioning dimension switching module switches to one-dimensional positioning mode or two-dimensional positioning mode when the scene classification result is a line-of-sight scene, switches to two-dimensional positioning mode when the scene classification result is a non-line-of-sight scene, and switches to three-dimensional positioning mode when the scene classification result is a dense multipath scene.

[0012] Compared with the prior art, the present invention has the following advantages: 1. This invention utilizes only the multipath characteristics of the ultra-wideband positioning signal itself for scene recognition and dimension decision-making, without relying on any additional sensors (such as IMU, temperature and humidity sensors, etc.), and has the advantages of low hardware cost, low power consumption, and strong universality.

[0013] 2. This invention classifies positioning scenarios into three categories: line-of-sight scenarios, non-line-of-sight scenarios, and dense multipath scenarios. Compared with the binary classification of LOS / NLOS in the prior art, the recognition granularity is finer, which can provide more accurate information support for the refined decision-making of positioning strategies.

[0014] 3. This invention establishes a mapping rule for "line-of-sight scenarios to low-dimensional positioning (one-dimensional / two-dimensional), non-line-of-sight scenarios to two-dimensional positioning, and dense multipath scenarios to three-dimensional positioning." This rule breaks with the conventional thinking that "complex environments should be simplified by reducing dimensions." In dense multipath environments, higher-dimensional three-dimensional positioning is used to utilize more base station information for redundant calculations to combat multipath errors; in line-of-sight environments, lower-dimensional positioning is used to save resources, achieving a dynamic optimal balance between positioning accuracy and system resource consumption.

[0015] 4. The present invention adopts a weighted fusion smooth switching strategy during the positioning dimension switching process, which avoids positioning jumps caused by abrupt changes in dimensions and ensures the continuity and stability of positioning results.

[0016] 5. This invention continuously monitors positioning accuracy and stability through a switching effect evaluation module, and automatically triggers re-identification and switching when performance degrades, forming a complete closed-loop adaptive system. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an ultra-wideband positioning dimension adaptive switching method based on signal multipath feature recognition according to an embodiment of the present invention. Figure 2This is a schematic diagram of an ultra-wideband positioning dimension adaptive switching system based on signal multipath feature recognition, according to an embodiment of the present invention. Detailed Implementation

[0019] 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.

[0020] like Figure 1 As shown, this invention provides an ultra-wideband positioning dimension adaptive switching method based on signal multipath feature recognition, comprising the following steps: Step S1: Receive the ultra-wideband positioning signal transmitted by the positioning target through an ultra-wideband positioning receiver, obtain channel impulse response (CIR) data, and extract multipath signal components from the channel impulse response data.

[0021] Specifically, an ultra-wideband (UWB) positioning system typically consists of positioning tags and multiple positioning anchors. The positioning tags emit UWB pulse signals, which are received by the anchors. Because UWB signals have extremely narrow pulse widths (nanosecond-level), their time resolution is extremely high, enabling them to distinguish multiple propagation paths with arrival time differences on the nanosecond level.

[0022] The signal received by the positioning base station can be represented as the convolution of the transmitted signal and the channel impulse response: ; in, In order to receive signals, For transmitting ultra-wideband pulse signals, For channel impulse response, It is additive white Gaussian noise. This represents the convolution operation.

[0023] Channel impulse response This reflects the multipath characteristics of a signal during propagation, and can be represented as the superposition of multiple propagation paths: ; in, The total number of multipath components. For the first The magnitude of the path (can be positive or negative, depending on the reflection coefficient). For the first Arrival time of each path Let be the Dirac impulse function.

[0024] In practical systems, by receiving signals By performing correlation operations with the local template signal or using time-domain post-processing algorithms such as CLEAN, an estimate of the channel impulse response can be extracted from the received signal. The channel impulse response is highly sparsity in the time domain, with the vast majority of sampling points having zero or near-zero energy, and significant non-zero energy only present at the actual multipath arrival time.

[0025] Extracting multipath signal components from channel impulse response data specifically includes the following sub-steps: (1) Noise threshold detection: Noise threshold detection is performed on the channel impulse response data, and the noise threshold is set to be lower than the preset noise threshold. The components are treated as noise and removed. Noise threshold. It is usually determined based on the noise floor of the received signal, and can be set to a certain multiple of the noise floor power (such as 3 times or 5 times) to achieve a balance between the detection probability and the false alarm probability.

[0026] (2) Peak detection: For peak values ​​exceeding the preset noise threshold Peak detection is performed on the components to identify the arrival time of each multipath component. and amplitude value Peak detection can employ a local maximum search method, which involves finding local maxima points in the channel impulse response sequence whose amplitudes are greater than those of adjacent sampling points.

[0027] (3) Feature calculation: based on the arrival time of each multipath component and amplitude value Calculate the multipath characteristic parameters required for subsequent steps.

[0028] Step S2: Perform time-domain and frequency-domain analysis on the extracted multipath signal components to extract multipath feature parameters characterizing the multipath propagation characteristics of the current positioning environment. The multipath feature parameters include at least the first path amplitude, first path arrival time, multipath delay spread, number of multipath components, and multipath power delay distribution. Specifically, this step extracts the following multipath feature parameters: (1) Initial diameter amplitude The first path refers to the first arriving component of the received signal whose amplitude exceeds the noise threshold. First path amplitude. This is the amplitude value of the path. In line-of-sight environments, the first path is usually a direct path with a large amplitude; in non-line-of-sight environments, the first path is a transmission or diffraction path after passing through an obstruction, and its amplitude is significantly attenuated; in dense multipath environments, the first path may be submerged in subsequent multipath components, making it difficult to identify accurately.

[0029] (2) Time of arrival of the first path The arrival time of the first path component. The arrival time of the first path is directly related to the actual distance from the positioning target to the positioning base station and is the basis for ranging and positioning calculations.

[0030] (3) Multipath delay spread Multipath delay spread is an important parameter for measuring the degree of channel time dispersion, and is defined as the square root of the second-order central moment of the multipath power delay distribution. ; in, For the first Arrival time of multipath components, For the first Power of multipath components, Mean Excess Delay is defined as: ; Multipath delay spread It reflects the degree of dispersion of multipath signals along the time axis. A larger value indicates a more dispersed temporal distribution of multipath components and a stronger frequency selectivity of the channel. In a line-of-sight environment, Typically smaller; in dense multipath environments. Significantly increased.

[0031] (4) Number of multipath components : This refers to an amplitude exceeding the noise threshold. The total number of multipath components. The number of multipath components directly reflects the complexity of the current environment. There are fewer multipath components in line-of-sight environments, and more multipath components in dense multipath environments.

[0032] (5) Multipath Power Delay Profile (PDP): The multipath power delay profile describes the distribution of received signal power over the arrival time delay. For discrete multipath components, the power delay profile can be expressed as: ; in, For the first Power of the multipath component.

[0033] Multipath power delay distribution is the basis for characterizing the characteristics of multipath channels, from which a variety of statistical characteristic parameters can be derived.

[0034] (6) Power ratio of first diameter The first-diameter power ratio is the ratio of the power of the first-diameter component to the total power of all multi-diameter components. ; in, The power is the first diameter component.

[0035] Head-to-diameter power ratio This reflects the proportion of the first path in the total received signal energy. In line-of-sight environments, the direct path carries the majority of the signal energy. The value is close to 1; in non-line-of-sight environments, the initial diameter is attenuated by occlusion. The value is significantly reduced; in dense multipath environments, signal energy is dispersed across numerous reflection paths. The value is usually very low.

[0036] Step S3: Input the extracted multipath feature parameters into the pre-trained scene classification model. The scene classification model identifies and classifies the current localized scene and outputs the scene classification result. The scene classification result includes at least line-of-sight (LOS) scenes, non-line-of-sight (NLOS) scenes, and dense multipath scenes.

[0037] In one alternative implementation, the scene classification model is a machine learning-based classification model, which is pre-trained through the following steps: (1) Sample collection: Collect ultra-wideband positioning signal samples under different positioning scenarios. Different positioning scenarios include at least line-of-sight scenarios, non-line-of-sight scenarios, and dense multipath scenarios. Specifically, UWB signal data can be collected in typical indoor environments (such as offices, warehouses, corridors, factory workshops, etc.) under line-of-sight conditions, non-line-of-sight conditions with obstructions, and dense multipath conditions with a large number of reflective objects.

[0038] (2) Feature extraction: Extract the multipath feature parameters of each signal sample as the training sample feature vector. The feature vector of each sample can be expressed as: ; Among them, feature vector The dimension depends on the number of multipath feature parameters extracted.

[0039] (3) Labeling: Label each signal sample with the corresponding scene category label. .

[0040] (4) Model training: The feature vectors of the training samples are... and corresponding scene category tags The input is fed into the initial classification model for training to obtain the scene classification model.

[0041] Scene classification models can be implemented using any of the following machine learning algorithms: Support Vector Machine (SVM), Random Forest, or Deep Neural Network (DNN).

[0042] Taking support vector machines as an example, the scene classification model obtains the optimal classification hyperplane by solving the following optimization problem: ; ; ; in, The normal vector of the classification hyperplane, For bias terms, A kernel function that maps input features to a high-dimensional space. For regularization parameters, As slack variables, This represents the total number of training samples.

[0043] For multi-classification problems (three-class scenarios), multiple binary classifiers can be constructed using a "one-vs-one" or "one-vs-rest" strategy.

[0044] During the online recognition phase, multipath feature parameters extracted in real time will be... Input a pre-trained scene classification model, and the model outputs the predicted scene category: ; in, To in a given feature vector Under the condition that the scene category is The posterior probability.

[0045] Step S4: Based on the scene classification results, determine the target positioning dimension that matches the current positioning scene. Positioning dimensions include one-dimensional positioning mode, two-dimensional positioning mode, and three-dimensional positioning mode.

[0046] The core of this step lies in establishing mapping rules from "scene category to positioning dimension". Specifically: When the scene classification result is a line-of-sight scene, the target positioning dimension is determined to be either a one-dimensional positioning mode or a two-dimensional positioning mode. When the scene classification result is a non-line-of-sight scene, the target positioning dimension is determined to be a two-dimensional positioning mode; When the scene classification result is a dense multipath scene, the target localization dimension is determined to be a three-dimensional localization mode.

[0047] The design principle of the above mapping rule is as follows: In line-of-sight scenarios, the first path (direct path) signal strength is high, and multipath components are few, making the positioning environment relatively simple. In this case, low-dimensional positioning (one-dimensional or two-dimensional) can meet the accuracy requirements while significantly reducing system resource consumption (reducing the number of required positioning base stations, communication overhead, and computational complexity). One-dimensional positioning is suitable for positioning along linear motion scenarios such as tracks and conveyor belts; two-dimensional positioning is suitable for positioning in ordinary indoor planar areas.

[0048] In non-line-of-sight scenarios, the initial path is attenuated by occlusion but is still identifiable. Multipath components increase but are not severe enough to overwhelm the initial path. In this case, a two-dimensional positioning mode is adopted, using information from at least three positioning base stations to calculate two-dimensional coordinates. The redundancy of information from multiple base stations can, to some extent, offset the ranging error caused by non-line-of-sight.

[0049] In dense multipath scenarios, numerous reflection paths overlap, potentially causing severe attenuation or even complete loss of identification of the initial path signal, resulting in low reliability of single ranging information. In such cases, a 3D positioning mode is employed, utilizing information from at least four positioning base stations for 3D spatial calculation. By introducing redundant measurement information from more base stations, the impact of multipath errors on positioning results can be effectively suppressed through optimization methods such as least squares in dense multipath environments. This "dimensional elevation in complex environments" design concept is one of the core innovations of this invention. While traditional understanding dictates that complex environments should be simplified, this invention takes the opposite approach, employing higher-dimensional positioning in dense multipath environments to acquire more redundant information to combat errors.

[0050] In an alternative implementation, system resource constraints are also considered in step S4: Obtain the number of currently available positioning base stations for the ultra-wideband positioning system. .

[0051] When the number of available positioning base stations is less than the minimum number of base stations required for the target positioning dimension, the target positioning dimension is downgraded to the highest positioning dimension supported by the number of available positioning base stations.

[0052] Specifically, one-dimensional positioning mode requires at least one positioning base station (based on prior constraints, such as the known target moving along a straight line); two-dimensional positioning mode requires at least three positioning base stations; and three-dimensional positioning mode requires at least four positioning base stations. The number of currently available positioning base stations... but If so, the 3D positioning mode will be downgraded to a 2D positioning mode; if but If so, it will be downgraded to a one-dimensional positioning mode.

[0053] Step S5: Switch the ultra-wideband positioning system from the current positioning dimension to the target positioning dimension, and use the positioning solution algorithm corresponding to the target positioning dimension to calculate the position of the positioning target.

[0054] In one optional implementation, in step S5, the positioning calculation algorithm corresponding to each positioning dimension includes: One-dimensional positioning mode: Employs a ranging and intersection algorithm based on Time of Arrival (TOA). Assuming the target moves along a known straight line (such as a track), the distance information between a single positioning base station and the target, combined with track geometric constraints, is used to calculate the target's one-dimensional coordinates on the track. Specifically, the coordinates of the positioning base station are set as follows: The unit vector of the orbital direction is The coordinates of a reference point on the track are Then the one-dimensional coordinates of the target satisfy: ; in, The coordinate vector for locating the base station. For the speed of light ( m / s This refers to the arrival time of the signal from the target location to the base station. This represents the Euclidean norm of a vector.

[0055] Two-dimensional positioning mode: Employs either a hyperbolic intersection algorithm based on Time Difference of Arrival (TDOA) or an intersection algorithm based on Angle of Arrival (AOA). Taking the TDOA algorithm as an example, the target coordinates are set as follows: , No. The coordinates of each positioning base station are Locate the target to the The time difference between each base station and the reference base station (e.g., base station 1) is Then we have: ; ; in, The number of base stations participating in the positioning ( The above system of equations is a hyperbolic system of equations, which can be solved using classic TDOA localization algorithms such as the Chan algorithm, the Fang algorithm, or the Taylor series expansion method.

[0056] 3D positioning mode: Employs either a spherical intersection algorithm based on Time of Arrival (TOA) or a hyperboloidal intersection algorithm based on Time Difference of Arrival (TDOA). Taking the TOA spherical intersection algorithm as an example, the 3D coordinates of the positioning target are set as follows: , No. The coordinates of each positioning base station are Locate the target to the The distance between the base stations is Then we have: ; in, Number of base stations participating in positioning The above system of equations is a spherical system of equations, which can be solved using the least squares method or Taylor series expansion. When measurement errors exist, nonlinear least squares optimization is typically employed. ; Preferably, this embodiment employs a smooth switching strategy during the positioning dimension switching process in step S5. During the transition from the current positioning dimension to the target positioning dimension, the positioning calculation algorithm for the current positioning dimension and the positioning calculation algorithm for the target positioning dimension are run simultaneously. The positioning results of the two positioning calculation algorithms are weighted and fused according to a preset weight coefficient, and the weight coefficient is gradually transitioned from the current positioning dimension to the target positioning dimension.

[0057] Specifically, let the positioning result of the current positioning dimension be... The positioning result of the target positioning dimension is Then during the handover transition period, the first At any given moment, the fused positioning result is: ; in, The weighting coefficients vary over time and satisfy the following conditions: ,and It monotonically increases from 0 to 1. A linear increment method can be used. Alternatively, a smoother increasing method, such as an S-curve, can be used. This is the preset transition time length.

[0058] By employing the smooth switching strategy described above, we can avoid location jumps caused by dimensional abrupt changes, thus ensuring the continuity and stability of the location results.

[0059] Preferably, after switching to the target positioning dimension, the positioning accuracy and stability of the positioning results are continuously monitored. Positioning accuracy can be evaluated using the root mean square error (RMSE) between the positioning results and the reference trajectory (or a known reference point): ; in, For the first The positioning results at each moment. For the first The actual position at each moment (which can be obtained through a high-precision reference system). This represents the number of sampling points.

[0060] Positioning stability can be assessed by the variance of the positioning results or the magnitude of jumps in positioning results between adjacent time steps: ; When the positioning accuracy (RMSE) or the positioning stability (Jitter) is lower than a preset threshold, the multipath signal acquisition step is re-executed to the positioning dimension switching step to achieve closed-loop adaptive adjustment.

[0061] like Figure 2 As shown, the present invention also provides an ultra-wideband positioning dimension adaptive switching system based on signal multipath feature recognition, which is used to implement the above method. The system includes the following modules: The ultra-wideband signal receiving module is used to receive the ultra-wideband positioning signal transmitted by the positioning target through the ultra-wideband positioning receiver and obtain channel impulse response data.

[0062] The ultra-wideband signal receiving module includes an ultra-wideband antenna, an RF front-end circuit, and a baseband signal processing unit. The ultra-wideband antenna receives nanosecond-level pulse signals transmitted by the positioning tag. The RF front-end circuit performs low-noise amplification, filtering, and down-conversion processing on the received signal. The baseband signal processing unit samples and digitizes the down-converted signal, and extracts the channel impulse response through correlation operations or the CLEAN algorithm. .

[0063] The multipath signal extraction module is used to extract multipath signal components from the channel impulse response data.

[0064] Specifically, the multipath signal extraction module includes: The noise threshold detection unit is used to perform noise threshold detection on the channel impulse response data, and to detect noise levels below a preset noise threshold. The components are removed as noise.

[0065] Peak detection unit, used to detect noise levels exceeding a preset noise threshold. Peak detection is performed on the components to identify the arrival time of each multipath component. and amplitude value .

[0066] The feature calculation unit is used to calculate the arrival time of each multipath component. and amplitude value Calculate multipath delay spread and multipath power delay distribution .

[0067] The multipath feature extraction module performs time-domain and frequency-domain analysis on the extracted multipath signal components to extract multipath feature parameters characterizing the multipath propagation characteristics of the current positioning environment. These multipath feature parameters include at least the amplitude of the first path. Time of arrival of the first path Multipath delay spread Number of multipath components Multipath power delay distribution and the head-to-diameter power ratio .

[0068] The scene classification module stores pre-trained scene classification models, inputs extracted multipath feature parameters into these models, and then classifies the currently located scene, outputting the scene classification result. The scene classification result includes at least line-of-sight scenes, non-line-of-sight scenes, and dense multipath scenes.

[0069] Specifically, the scene classification model in the scene classification module is a classification model pre-trained using any one of the machine learning algorithms, such as support vector machine, random forest, or deep neural network.

[0070] Taking deep neural networks as an example, a scene classification model can include an input layer, multiple hidden layers, and an output layer. The number of neurons in the input layer is equal to the number of multipath feature parameters, and the output layer uses the Softmax activation function to output the probability distribution of three scene categories. ; in, and The first The weight vector and bias term corresponding to each category.

[0071] The positioning dimension decision module determines the target positioning dimension that matches the current positioning scene based on the scene classification results. Positioning dimensions include one-dimensional positioning mode, two-dimensional positioning mode, and three-dimensional positioning mode. Table 1 shows the module's built-in scene-dimensional mapping table.

[0072] Table 1

[0073] The positioning dimension decision module also receives the number of available positioning base stations from the base station number monitoring module. ,exist If the number of base stations is less than the minimum number required for the target positioning dimension, dimension downgrading is performed.

[0074] The positioning dimension switching module is used to switch the ultra-wideband positioning system from the current positioning dimension to the target positioning dimension.

[0075] Specifically, the positioning dimension switching module adopts a smooth switching strategy. During the transition from the current positioning dimension to the target positioning dimension, the positioning calculation algorithm for the current positioning dimension and the positioning calculation algorithm for the target positioning dimension are run simultaneously, according to preset weight coefficients. The positioning results from the two positioning algorithms are weighted and fused, and the weight coefficients are gradually transitioned from the current positioning dimension to the target positioning dimension: ; The positioning calculation module is used to calculate the position of the positioning target using a positioning calculation algorithm corresponding to the positioning dimension of the target.

[0076] Specifically, the positioning solution module supports three positioning solution modes: One-dimensional positioning solution: A TOA-based ranging and intersection algorithm is used, combined with orbital geometric constraints, to solve for the one-dimensional coordinates. ; Two-dimensional positioning solution: The two-dimensional coordinates are solved using either a hyperbolic intersection algorithm based on TDOA or an intersection algorithm based on AOA. ; 3D positioning calculation: The 3D coordinates are solved using either a TOA-based spherical intersection algorithm or a TDOA-based hyperboloid intersection algorithm. .

[0077] In one optional embodiment, the system of the present invention further includes: The base station quantity monitoring module is used to obtain the number of currently available positioning base stations in the ultra-wideband positioning system. This module monitors the communication status of each positioning base station in the system in real time, counts the number of base stations currently in normal working condition (signal valid, clock synchronized), and... Provided to the positioning dimension decision module.

[0078] The positioning dimension decision module considers the number of available positioning base stations. If the number of base stations required for the target positioning dimension is less than the minimum number required for the target positioning dimension, the target positioning dimension will be downgraded to the highest positioning dimension supported by the number of available positioning base stations.

[0079] In one optional embodiment, the system of the present invention further includes: The switching effect evaluation module is used to continuously monitor the positioning accuracy and positioning stability of the positioning results after switching to the target positioning dimension, and to trigger the multipath signal extraction module to re-execute multipath signal extraction and subsequent operations when the positioning accuracy or positioning stability is lower than a preset threshold.

[0080] Specifically, the switching effect evaluation module calculates the RMSE and Jitter metrics of the positioning results in real time: ; ; when or Time (of which) and These are preset accuracy thresholds and stability thresholds, respectively, which trigger the system to re-execute the entire process of multipath signal extraction, multipath feature extraction, scene recognition, dimension decision-making, and dimension switching to achieve closed-loop adaptive adjustment.

[0081] The following example, using a typical indoor warehouse positioning scenario, illustrates the specific application process of this invention.

[0082] Scenario: Six UWB positioning base stations are deployed in a large warehouse, distributed at the four corners and the center. An AGV (Automated Guided Vehicle) equipped with a UWB positioning tag needs to navigate autonomously within the warehouse.

[0083] Initial state: The AGV is located in the open area in the center of the warehouse. At this time, the line of sight is good and the multipath component is small.

[0084] System operation process: 1. The UWB tags on the AGV periodically transmit positioning pulses, and each base station receives the signals and extracts CIR data.

[0085] 2. The multipath feature extraction module extracts the first diameter amplitude from the CIR. (Normalized value), multipath delay spread ns, number of multipath components Power ratio of first diameter .

[0086] 3. The scene classification model takes these features as input and outputs the scene classification result as "line-of-sight scene" (confidence level 92%).

[0087] 4. The positioning dimension decision module determines the target positioning dimension as "two-dimensional positioning mode" according to the mapping rules (the warehouse is a planar area and does not require three-dimensional positioning).

[0088] 5. The system switches to two-dimensional positioning mode, using the three nearest base stations to perform TDOA hyperbolic intersection positioning, achieving a positioning accuracy within 10cm. The system has low power consumption, requiring only communication with the three base stations.

[0089] Scenario change: When the AGV travels to an area with dense shelving, the signal is reflected and blocked multiple times by the metal shelving.

[0090] System response: 1. CIR data received by the base station shows that the first diameter amplitude has decreased to The multipath delay spread increases to ns, the number of multipath components increases to The head-to-diameter power ratio decreased to .

[0091] 2. The scene classification model outputs "dense multipath scene" (confidence level 88%).

[0092] 3. The positioning dimension decision module determines the target positioning dimension as "three-dimensional positioning mode".

[0093] 4. The system smoothly switches to 3D positioning mode, utilizing 5 available base stations for TOA spherical intersection positioning. Although the ranging of a single base station may be affected by multipath errors, the redundant information from multiple base stations is used for least squares optimization, effectively suppressing multipath errors and maintaining positioning accuracy at around 25cm.

[0094] Closed-loop feedback: The switching effect evaluation module detected an RMSE of 22cm in 3D positioning mode, which is better than the preset threshold of 30cm. The system maintains the current positioning dimension unchanged. If the environment deteriorates further and causes the RMSE to exceed the threshold, the system will automatically trigger a re-identification and switching process.

[0095] Through the above process, the present invention achieves a dynamic optimal balance between positioning accuracy and system resource consumption, saving resources in simple environments and ensuring accuracy in complex environments.

[0096] This invention can be widely applied to various ultra-wideband positioning systems, and is especially suitable for high-precision positioning applications in complex indoor environments, including but not limited to: Industrial automation: AGV / AMR navigation and positioning, production line material tracking; Smart warehousing: cargo positioning and forklift dispatching; Public safety: Location of fire and rescue personnel, location of personnel in mines; Smart buildings: indoor navigation, people flow monitoring; Internet of Things (IoT): Asset tracking, smart homes.

[0097] This invention does not rely on any additional sensors and can achieve adaptive positioning using only the UWB signal itself. It has outstanding advantages such as low hardware modification cost, simple deployment, and wide applicability, and has good industrial practicality and market promotion value.

[0098] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for adaptive switching of ultra-wideband positioning dimensions based on signal multipath feature recognition, characterized in that, Includes the following steps: S1. Receive the ultra-wideband positioning signal transmitted by the positioning target through an ultra-wideband positioning receiver, obtain channel impulse response data, and extract multipath signal components from the channel impulse response data; S2. Perform time-domain and frequency-domain analysis on the extracted multipath signal components to extract multipath feature parameters that characterize the multipath propagation characteristics of the current positioning environment. The multipath feature parameters include at least the first path amplitude, the first path arrival time, the multipath delay spread, the number of multipath components, and the multipath power delay distribution. S3. Input the extracted multipath feature parameters into the pre-trained scene classification model, and the scene classification model identifies and classifies the current location scene, and outputs the scene classification result. The scene classification result includes at least line-of-sight scene, non-line-of-sight scene and dense multipath scene. S4. Based on the scene classification results, determine the target positioning dimension that matches the current positioning scene. The positioning dimension includes one-dimensional positioning mode, two-dimensional positioning mode and three-dimensional positioning mode. S5. Switch the ultra-wideband positioning system from the current positioning dimension to the target positioning dimension, and use the positioning calculation algorithm corresponding to the target positioning dimension to calculate the position of the positioning target; In step S4, when the scene classification result is a line-of-sight scene, the system switches to a one-dimensional positioning mode or a two-dimensional positioning mode; when the scene classification result is a non-line-of-sight scene, the system switches to a two-dimensional positioning mode; and when the scene classification result is a dense multipath scene, the system switches to a three-dimensional positioning mode.

2. The method according to claim 1, characterized in that, In S2, the multipath characteristic parameter also includes the first-path power ratio, which is the ratio of the power of the first-path component to the total power of all multipath components.

3. The method according to claim 1, characterized in that, The scene classification model is a machine learning-based classification model, which is pre-trained through the following steps: Collect ultra-wideband positioning signal samples under different positioning scenarios, including at least line-of-sight scenarios, non-line-of-sight scenarios, and dense multipath scenarios. Extract the multipath feature parameters of each signal sample as the training sample feature vector; Label each signal sample with its corresponding scene category. The training sample feature vectors and corresponding scene category labels are input into the initial classification model for training to obtain the scene classification model.

4. The method according to claim 1, characterized in that, In step S4, determining the target positioning dimension that matches the current positioning scene further includes: Obtain the number of currently available positioning base stations for the ultra-wideband positioning system; When the number of available positioning base stations is less than the minimum number of base stations required for the target positioning dimension, the target positioning dimension is downgraded to the highest positioning dimension supported by the number of available positioning base stations.

5. The method according to claim 1, characterized in that, In step S5, the positioning calculation algorithms corresponding to each positioning dimension include: The one-dimensional positioning mode uses a time-of-arrival ranging and intersection algorithm; The two-dimensional positioning mode employs either a hyperbolic intersection algorithm based on the time difference of arrival or an intersection algorithm based on the angle of arrival. The three-dimensional positioning mode employs either a spherical intersection algorithm based on arrival time or a hyperboloid intersection algorithm based on the time difference of arrival.

6. An ultra-wideband positioning dimension adaptive switching system based on signal multipath feature recognition, characterized in that, include: The ultra-wideband signal receiving module is used to receive the ultra-wideband positioning signal transmitted by the positioning target through the ultra-wideband positioning receiver and obtain channel impulse response data; A multipath signal extraction module is used to extract multipath signal components from the channel impulse response data; The multipath feature extraction module is used to perform time-domain and frequency-domain analysis on the extracted multipath signal components and extract multipath feature parameters to characterize the multipath propagation characteristics of the current positioning environment. The multipath feature parameters include at least the first path amplitude, the first path arrival time, the multipath delay spread, the number of multipath components, and the multipath power delay distribution. The scene classification module is used to store pre-trained scene classification models, input the extracted multipath feature parameters into the scene classification models, and have the scene classification models identify and classify the current location scene and output scene classification results. The scene classification results include at least line-of-sight scenes, non-line-of-sight scenes and dense multipath scenes. The positioning dimension decision module is used to determine the target positioning dimension that matches the current positioning scene based on the scene classification results. The positioning dimension includes one-dimensional positioning mode, two-dimensional positioning mode and three-dimensional positioning mode. The positioning dimension switching module is used to switch the ultra-wideband positioning system from the current positioning dimension to the target positioning dimension; The positioning calculation module is used to calculate the position of the positioning target using a positioning calculation algorithm corresponding to the positioning dimension of the target. Specifically, the positioning dimension decision module switches to one-dimensional positioning mode or two-dimensional positioning mode when the scene classification result is a line-of-sight scene, switches to two-dimensional positioning mode when the scene classification result is a non-line-of-sight scene, and switches to three-dimensional positioning mode when the scene classification result is a dense multipath scene.

7. The system according to claim 6, characterized in that, Also includes: The base station number monitoring module is used to obtain the number of currently available positioning base stations in the ultra-wideband positioning system; When the number of available positioning base stations is less than the minimum number of base stations required for the target positioning dimension, the positioning dimension decision module downgrades the target positioning dimension to the highest positioning dimension supported by the number of available positioning base stations.

8. The system according to claim 6, characterized in that, Also includes: The switching effect evaluation module is used to continuously monitor the positioning accuracy and positioning stability of the positioning results after switching to the target positioning dimension, and to trigger the multipath signal extraction module to re-execute multipath signal extraction and subsequent operations when the positioning accuracy or positioning stability is lower than a preset threshold.

9. The system according to claim 6, characterized in that, The positioning dimension switching module adopts a smooth switching strategy. During the transition from the current positioning dimension to the target positioning dimension, the positioning solution algorithm of the current positioning dimension and the positioning solution algorithm of the target positioning dimension are run simultaneously. The positioning results of the two positioning solution algorithms are weighted and fused according to a preset weight coefficient, and the weight coefficient is gradually transitioned from the current positioning dimension to the target positioning dimension.

10. The system according to claim 6, characterized in that, The multipath signal extraction module includes: The noise threshold detection unit is used to perform noise threshold detection on the channel impulse response data and discard components with amplitudes lower than a preset noise threshold as noise. A peak detection unit is used to detect the peak values ​​of components that are higher than the preset noise threshold and to identify the arrival time and amplitude value of each multipath component. The feature calculation unit is used to calculate the multipath delay spread and the multipath power delay distribution based on the arrival time and amplitude value of each multipath component.

Images