A UWB Multi-Dimensional Fusion Positioning Method and System

Abstract

The present invention relates to the technical field of wireless communication and positioning, and particularly relates to a UWB multi-dimensional fusion positioning method and system. The present invention proposes the following solutions: obtaining timestamp data from UWB tags to base stations, and screening available base stations based on the Z-count normalization method; calculating two-dimensional coordinates by the three-base-station combination method, and eliminating abnormal data by combining the position information of the previous epoch; calculating one-dimensional coordinates by using two base stations, or determining the zero-dimensional position based on the base station with the minimum timestamp; finally, fusing multi-dimensional coordinates and optimizing the tag position by combining historical trajectories. This method can adapt to non-line-of-sight environments, reduce positioning errors, improve stability during dimension conversion, and is applicable to high-precision indoor positioning scenarios such as smart factories, intelligent warehouses, and smart healthcare.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication and positioning, and particularly to a UWB multi-dimensional fusion positioning method and system. Background Art

[0002] Indoor positioning has gradually become a research hotspot in the field of navigation and positioning, and accurate and real-time indoor location services have become an indispensable important information resource. Like GPS, indoor positioning is affected by non-line-of-sight and its accuracy is reduced, and current indoor positioning is mostly two-dimensional with a single positioning dimension.

[0003] For example, the Chinese patent with the authorization announcement number CN112702699B discloses an indoor positioning method integrating UWB and LiDAR, including: deploying UWB and LiDAR for relevant measurements; obtaining the input of the UWB sensor and calculating the approximate positioning information of UWB; obtaining the input of LiDAR and obtaining the approximate positioning information of LiDAR; calculating coordinate transformation to convert the positioning information of different sensors to the same coordinate system; fitting the motion states of the two to obtain the orientation information of the UWB positioning and complement the positioning information of UWB; judging positioning anomalies; performing repositioning in case of positioning anomalies; and outputting real-time positioning results.

[0004] For example, the patent application with the publication number CN117412372A discloses a UWB positioning device and a UWB positioning method. The UWB positioning device includes a first UWB antenna combination and a first adjustment mechanism, and the first UWB antenna combination includes a first UWB antenna and a second UWB antenna; the first adjustment mechanism is used to adjust the center distance between the first UWB antenna and the second UWB antenna.

[0005] The above technical solutions all have the problems proposed in this background art: dimension selection errors and reduced position accuracy are likely to occur in the intersection area of different dimensions, and affected by non-line-of-sight, it will also lead to a decrease in accuracy or even no position solution. To solve the above problems, this application designs a UWB multi-dimensional fusion positioning method and system. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a UWB multi-dimensional fusion positioning method and system in view of the deficiencies of the prior art, which obtains the timestamp data from the UWB tag to the base station, and filters available base stations based on the Z-count normalization method; calculates the two-dimensional coordinates by the three-base-station combination method, and eliminates abnormal data in combination with the position information of the previous epoch; calculates the one-dimensional coordinates by using two base stations, or determines the zero-dimensional position based on the base station with the smallest timestamp; finally, fuses the multi-dimensional coordinates and optimizes the tag position in combination with the historical trajectory. This method can adapt to the non-line-of-sight environment, reduce the positioning error, improve the stability during dimension conversion, and is applicable to high-precision indoor positioning scenarios such as smart factories, intelligent warehouses, and smart healthcare.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A UWB multi-dimensional fusion positioning method, which is used for indoor positioning of an object to be positioned, wherein a UWB tag is configured on the object to be positioned, and the UWB multi-dimensional fusion positioning method includes:

[0009] Obtain the positioning data from the UWB tag of each object to be positioned to each base station and the position information of the previous epoch.

[0010] Determine the available base stations among all base stations according to the positioning data.

[0011] Input the position information of the previous epoch into a preset proximity model, and the proximity model outputs the tag position of the current epoch, and the tag position of the current epoch is calculated according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates of the UWB tag.

[0012] Fuse the current epoch tag position with the current position prediction information obtained from the position information of the previous epoch to obtain the updated position of the current epoch.

[0013] The positioning data includes the timestamp data from the UWB tag to multiple base stations, the base station coordinates, and the dimension type identification information.

[0014] The determining the available base stations among all base stations according to the positioning data includes:

[0015] Obtain the timestamp data from the UWB tag to each base station.

[0016] Calculate the normalized values of the timestamps of each base station according to the Z-count normalization method, and eliminate abnormal base stations greater than or equal to a preset standard threshold.

[0017] Sort the base stations according to the timestamp size, and determine the sorting index of different dimension base stations according to the dimension type identification information of the base stations, and output the available base stations.

[0018] The output of the current epoch tag position includes:

[0019] According to the timestamp sorting index of the available base stations, select combinations of multiple two-dimensional base stations, and calculate multiple groups of two-dimensional position coordinates according to the three-base-station combination method;

[0020] According to the previous epoch position information, eliminate the solution results in the multiple groups of two-dimensional position coordinates whose distance from the previous epoch position exceeds a preset distance threshold, and use the mean value of the remaining two-dimensional position coordinates as the two-dimensional position coordinates of the current epoch;

[0021] According to the dimension type identifier of the available base stations, select combinations of multiple one-dimensional base stations, and select the first valid solution result as the one-dimensional position coordinates of the current epoch according to the two-base-station combination method;

[0022] Select the coordinates of the zero-dimensional base station with the smallest timestamp as the zero-dimensional position coordinates of the current epoch;

[0023] Calculate the current epoch tag position according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates.

[0024] The calculation of the current epoch tag position according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates includes:

[0025] If the current epoch is the first epoch, use the mean value of the two-dimensional, one-dimensional, and zero-dimensional position coordinates as the current epoch tag position;

[0026] If the current epoch is not the first epoch, calculate the Euclidean distance between the position coordinates of each dimension and the previous epoch position, and select the dimension position with the closest Euclidean distance as the current epoch tag position.

[0027] The current position prediction information obtained according to the previous epoch position information includes:

[0028] Calculate the position information of the previous epoch and multiple previous epochs to determine the movement trajectory of the UWB tag;

[0029] According to the movement trajectory, calculate the movement speed of the UWB tag, and predict the simulated speed of the current epoch according to the time interval;

[0030] According to the simulated speed, calculate the current position prediction information through the state transition model.

[0031] The fusion of the current epoch tag position to obtain the updated position of the current epoch includes:

[0032] According to the current position prediction information and the current epoch tag position, calculate the adaptive speed robust filtering gain;

[0033] Calculate the updated position of the current epoch according to the adaptive speed robust filtering gain.

[0034] The UWB multi-dimensional fusion positioning method further includes:

[0035] Obtain real-time environmental parameters, and adjust the standard threshold according to the real-time environmental parameters;

[0036] Optimize the contribution ratio of the base stations in position calculation according to the historical signal stability and positioning accuracy of base stations in different dimensions.

[0037] A UWB multi-dimensional fusion positioning system, the system includes a data acquisition module, a base station screening module, a position calculation module and a position fusion module;

[0038] The data acquisition module is used to obtain the timestamp data, base station coordinates and base station dimension type identification information from the UWB tag to each base station, and store the relevant data;

[0039] The base station screening module is used to process the data obtained by the data acquisition module, eliminate abnormal base stations based on the Z-count normalization method, sort the base stations according to the timestamp size, and at the same time determine the sorting index of base stations in different dimensions according to the base station dimension type identification information, and output available base stations;

[0040] The position calculation module calculates the label position of the current epoch based on the data of the available base stations;

[0041] The position fusion module calculates the motion trajectory of the UWB tag based on the position information of the previous epoch and multiple previous epochs, calculates the current position prediction information using the state transition model, and fuses the label position of the current epoch and the current position prediction information;

[0042] The position calculation module includes:

[0043] The two-dimensional position calculation unit is used to select combinations of multiple two-dimensional base stations, calculate multiple groups of two-dimensional position coordinates according to the three-base station combination method, and use the mean value of the two-dimensional position coordinates as the two-dimensional position coordinates of the current epoch;

[0044] The one-dimensional position calculation unit is used to select combinations of multiple one-dimensional base stations, and select the first valid calculation result as the one-dimensional position coordinates of the current epoch according to the two-base station combination method;

[0045] The zero-dimensional position calculation unit is used to select the coordinates of the zero-dimensional base station with the smallest timestamp as the zero-dimensional position coordinates of the current epoch;

[0046] The label position calculation unit calculates the label position of the current epoch according to the two-dimensional position coordinates, one-dimensional position coordinates and zero-dimensional position coordinates.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] The present invention provides a UWB multi-dimensional fusion positioning method and system. Through Z-count standardization data screening, multi-dimensional fusion calculation, and historical trajectory optimization, the accuracy and stability of indoor positioning are significantly improved. Compared with traditional methods, the present invention can dynamically adapt to complex environments and still maintain high-precision positioning in non-line-of-sight (NLOS) conditions, crowded scenarios, multi-dimensional intersection areas, etc. Through the intelligent base station optimization strategy, the system can real-time screen the optimal base station combination, reduce the interference of low-quality signals, and at the same time combine two-dimensional, one-dimensional, and zero-dimensional multi-dimensional calculation methods to ensure the stability and continuity of the positioning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Other features, objects, and advantages of the present invention will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings:

[0050] Figure 1 It is a schematic flow chart of a UWB multi-dimensional fusion positioning method according to Embodiment 1 of the present invention;

[0051] Figure 2 It is a flow chart of a standard threshold adjustment method according to Embodiment 1 of the present invention;

[0052] Figure 3 It is a flow chart of a position calculation method based on the position of the previous epoch according to Embodiment 1 of the present invention;

[0053] Figure 4 It is a module diagram of a UWB multi-dimensional fusion positioning system according to Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0055] Embodiment 1

[0056] Please refer to Figure 1 , an embodiment provided by the present invention: a UWB multi-dimensional fusion positioning method, the UWB multi-dimensional fusion positioning method is used for indoor positioning of an object to be located, especially for precise positioning in a multi-dimensional environment intersection area and non-line-of-sight conditions, wherein a UWB tag is configured on the object to be located, and the specific steps of the UWB multi-dimensional fusion positioning method are as follows:

[0057] S1: Obtain the positioning data from the UWB tag to each base station and the position information of the previous epoch;

[0058] In this embodiment, by deploying multiple UWB base stations, the timestamp data of the UWB tag of the object to be located to each base station, the base station coordinate information, and the dimension type identifier are obtained in real time. At the same time, combined with the position information of the previous epoch, it provides basic data for subsequent calculations. The purpose of doing this is to improve the positioning accuracy, ensure that the position information of the tag can be continuously and effectively obtained in a complex environment. Even if there is signal occlusion or abnormal base station data, it can be compensated by the information of the previous epoch to improve data integrity;

[0059] S2: Determine the available base stations among all base stations according to the positioning data;

[0060] In this embodiment, in order to screen out the base stations that contribute to the positioning calculation, first, the Z-count normalization method is used to detect and eliminate the base stations with abnormal timestamps. Subsequently, the remaining base stations are sorted according to the timestamp size, and according to the dimension type information of the base stations, the sorting indexes of different dimension base stations are determined, and finally the available base stations are output. This method can effectively exclude the base stations with abnormal data, improve the accuracy of the positioning calculation, and optimize the selection of base stations, making the data used in the calculation more reliable and avoiding the positioning error caused by the interference of abnormal base stations;

[0061] S3: Input the position information of the previous epoch into a preset proximity model and output the tag position of the current epoch;

[0062] In this embodiment, in order to solve the computational complexity and non-line-of-sight influence of multi-dimensional positioning in an indoor environment, a proximity model is constructed. This model is based on the position information of the previous epoch and combines the sorting indexes of available base stations to calculate two-dimensional, one-dimensional, and zero-dimensional positioning results. First, the three-base-station combination method is used to calculate the two-dimensional position coordinates, and the solution results that are too far from the previous epoch are eliminated; secondly, in the one-dimensional scenario, the first effective two-base-station combination is selected for one-dimensional position calculation; for zero-dimensional positioning, the zero-dimensional base station with the smallest timestamp is selected as the zero-dimensional position. By this means, it can ensure high-precision seamless positioning in a complex indoor environment, avoid the positioning instability problem during dimension conversion, and improve the calculation accuracy of the tag position;

[0063] S4: According to the current position prediction information obtained from the position information of the previous epoch, fuse the tag position of the current epoch to obtain the updated position of the current epoch;

[0064] In this embodiment, in order to further improve the continuity of positioning and the anti-error ability, a current position prediction model is constructed. Using the tag position information of the previous epoch and multiple previous epochs, the movement trajectory and speed are calculated, and the theoretical position of the current epoch is predicted through the state transition model. Then, the tag position calculated in the current epoch is compared with the predicted position, the deviation value is calculated, and the two are fused through a dynamic weighting strategy to finally obtain the updated position of the optimized current epoch. If the calculation in the current epoch fails, the predicted position is directly used as the final position to ensure that the positioning is not interrupted and the robustness of the system in a complex environment is improved. This method can effectively reduce the positioning error caused by factors such as base station signal interference and obstacle occlusion, and improve the accuracy and stability of positioning.

[0065] Specifically, in the existing indoor positioning technologies, the traditional UWB positioning method is usually limited to two-dimensional plane positioning and is difficult to adapt to the complexity of multi-dimensional environments. Especially in special scenarios such as floor transition areas, vertical spaces (such as shelves, stairs, escalators), and non-line-of-sight (NLOS) environments, the traditional method is prone to problems such as error accumulation and unstable dimension conversion. For example, the two-dimensional positioning technology based on UWB usually only relies on the layout of plane base stations, and it is difficult to distinguish whether the object to be positioned is on the same plane or at different heights at different height levels. Further, when the positioning target moves to special scenarios such as elevators, stairs, and three-dimensional shelves, the existing methods are difficult to accurately track the movement trajectory of the target and make a seamless transition between different heights, resulting in a decrease in positioning accuracy or even loss of the target. In addition, in a non-line-of-sight environment, the UWB signal is easily affected by occlusion and reflection, resulting in a significant increase in ranging error and a large deviation in the calculation of the target position. Therefore, there are great challenges in the positioning robustness and high-precision continuous tracking of the existing UWB positioning method in different dimensional intersection areas and complex environments.

[0066] In this embodiment, a UWB multi-dimensional fusion positioning method is provided. Taking a shopping mall environment as an example, UWB base stations are deployed on multiple floors of the shopping mall, and additional base stations are arranged at key positions such as escalators, elevators, stairs, and commodity areas to enhance the positioning accuracy during dimension conversion. When consumers in the shopping mall wear UWB tags (such as worn on wristbands, smart shopping carts, or smart terminals), the system will real-time obtain the timestamp data of the UWB tags of the consumers to the base stations and calculate in combination with the position information of the previous epoch.

[0067] Exemplarily, in the traditional UWB two-dimensional positioning method, if a consumer takes an elevator from the first floor to the second floor of a shopping mall, due to the propagation characteristics of UWB signals, the signals may be received by the base stations on the first floor and the second floor at the same time. It is difficult for the traditional method to distinguish whether the consumer is still on the first floor or has reached the second floor, resulting in positioning errors. In this embodiment, by constructing a multi-dimensional tag position calculation model, available base stations are first screened based on timestamp data, and then a proximity calculation model is constructed by combining the position of the previous epoch, the base station height information, and the base station dimension type to automatically screen the optimal dimension of the current epoch, ensuring that the positioning information does not get confused when the height dimension of the consumer changes. It can accurately determine the floor where the consumer is located in scenarios such as shopping mall elevators, stairs, escalators, etc., and ensure seamless connection of the positioning information.

[0068] Exemplarily, in a shopping mall, factors such as large commodity shelves, dense crowds, and walls will block UWB signals, resulting in an increase in non-line-of-sight error (NLOS), thus affecting the positioning accuracy. Traditional UWB positioning methods usually have difficulty effectively eliminating the influence of NLOS, which may lead to excessive position errors of consumers and even the phenomenon of position jumps (i.e., sudden position changes). This embodiment adopts an adaptive speed robust filtering model based on the current position, which can analyze the tag movement trajectory and speed changes in real time and filter out abnormal data. For example, when a consumer walks to a crowded dining area and the UWB signal is blocked, there may be a short-term position offset. This method can predict the movement trajectory by combining the current position information and correct the tag position through an adaptive filtering algorithm to ensure high-precision positioning of the consumer in a non-line-of-sight environment.

[0069] Preferably, in this embodiment, the UWB multi-dimensional fusion positioning method of the present application is not only applicable to the shopping mall environment, but can also be widely applied to other scenarios that require high-precision indoor positioning, such as complex environments like smart factories, intelligent warehouses, hospital management, and underground parking lots.

[0070] Specifically, in a smart factory, this application can be used for equipment on an automated production line, AGVs (Automated Guided Vehicles), forklifts, and worker positioning management to ensure optimized material transportation routes, improve operation efficiency, and avoid safety hazards between workers and automated equipment. In an intelligent warehousing environment, this application can accurately identify the spatial positions of goods in a multi-layer shelf system. Even in the case of high-density stacking of goods, it can still maintain stable positioning accuracy and effectively improve inventory management capabilities. In hospital management, this application can be used to track the positions of medical equipment, patients, and medical staff in real time. For example, in key areas such as operating rooms and ICUs, it ensures that the required equipment and personnel can be quickly located in case of emergencies, improving the hospital's operation efficiency and medical safety. In addition, in non-line-of-sight environments such as underground parking lots where GPS signals cannot cover, this application can achieve precise vehicle guidance and parking space management, avoiding parking troubles caused by positioning errors and improving the intelligent management level of the parking lot.

[0071] The specific steps of S2 are as follows:

[0072] S2.1: Obtain the timestamp data of the UWB tag to each base station;

[0073] Specifically, the UWB base station receives signals from the tag and records the timestamp data of the tag to each base station. These timestamp data are generated based on the Time of Arrival (TOA) of the UWB signal and represent the time length from the transmission of the tag signal to its arrival at the base station. The purpose of obtaining the timestamp data is to provide the core raw data for subsequent positioning calculations. By collecting the timestamp information of all base stations, the spatio-temporal relationship between the base stations and the tag can be constructed to ensure that the positioning calculation has sufficient input data support. Especially in a complex indoor environment, obtaining data from multiple base stations can provide rich basic information for subsequent error screening and dimension fusion calculations, thereby effectively improving the stability and robustness of positioning.

[0074] S2.2: Calculate the standardized values of the timestamps of each base station according to the Z-count normalization method, and eliminate abnormal base stations with values greater than or equal to the preset standard threshold;

[0075] In this embodiment, in order to eliminate possible abnormal values (such as measurement errors caused by signal reflection and interference) in the timestamp data, the Z-count normalization method is used to normalize the timestamp data. First, calculate the Z-core value of the epoch base station timestamp, that is, the standardized value. The calculation formula of the standardized value is:

[0076] ;

[0077] Where, represents the timestamp of a single base station, Represents the timestamps of all base stations in the current epoch, Represents the average value of T, Represents the function for calculating the average value, Represents the standard deviation value of T, Represents the function for calculating the standard deviation, Represents the Z-count normalization value of each element in T, reflecting the degree of discretization of the elements in the dataset;

[0078] Specifically, the normalization calculation is performed by normalizing the mean and standard deviation of all base station timestamps to obtain the degree of discretization (i.e., the Z-core value) of each base station timestamp. The Z-core value can intuitively reflect the deviation degree of a base station timestamp from the overall data distribution;

[0079] In this embodiment, a gross error identification F is constructed based on the Z-core value of the epoch base station timestamp, and qualified base stations are selected according to the gross error identification F. The representation of the gross error identification F is as follows:

[0080] ;

[0081] An F value of 1 indicates that the base station is unqualified and is a gross error base station, and a value of 0 indicates that the base station is an available base station. B represents the standard threshold, and the set of available base stations is represented as ;

[0082] Specifically, when the Z-core value exceeds the preset standard threshold, it is considered that the data of this base station is an outlier, and this base station is directly excluded. It can effectively exclude invalid data caused by signal reflection or occlusion in a non-line-of-sight environment, improve the reliability of the data. It can ensure that subsequent positioning calculations are based on high-quality input data, reducing the possibility of error propagation and accumulation;

[0083] In this embodiment, the setting of the standard threshold can be obtained by those skilled in the art through a large number of experiments. However, in reality, the propagation characteristics of UWB signals will be affected by air temperature and humidity and wall materials, thereby affecting the ranging accuracy and the stability of base station timestamp data.

[0084] Specifically, in order to reduce the errors caused by environmental factors and improve the accuracy of UWB positioning, this application proposes a dynamic standard threshold adjustment method based on real-time environmental parameters. By real-time monitoring of temperature and humidity and building structure information, the standard threshold of the Z-count normalization method is optimized and adjusted to adapt to different environmental conditions and ensure the rationality of base station data screening;

[0085] Specifically, the absorption of UWB signals by water vapor in the air increases with the increase in humidity. Especially in high-humidity environments (such as the rainy season, indoor constant-humidity warehouses, and underground spaces), the propagation speed of UWB signals will change slightly, resulting in an increase in ranging errors. In addition, temperature changes may affect the working state of electronic components, leading to clock drift in the base station and further affecting the stability of timestamp data;

[0086] Specifically, different building materials have different penetrability for UWB signals. Especially in environments such as concrete, metal walls, and glass curtain walls, UWB signals attenuate significantly, easily forming non-line-of-sight (NLOS) propagation paths, resulting in abnormal timestamp data;

[0087] Please refer to Figure 2 , the flowchart of the standard threshold adjustment method in the embodiment of the present invention, aims to solve the ranging error problem caused by environmental factor changes during indoor UWB positioning, and improve the stability and accuracy of the system. When UWB signals propagate in a complex environment, affected by factors such as air temperature and humidity, wall materials, and signal attenuation characteristics, ranging data may deviate. If these environmental factors are not dynamically compensated, the traditional fixed threshold rejection method may lead to problems such as excessive errors or incorrect rejection of valid base stations. Therefore, this method adaptively adjusts the standard threshold in UWB positioning calculations through real-time environmental data collection and signal propagation characteristic analysis to optimize error rejection and positioning accuracy. The specific steps of the dynamic standard threshold adjustment method based on real-time environmental parameters are as follows:

[0088] SA1: Real-time collect environmental temperature and humidity data, and calculate the correction coefficient of the signal propagation rate in combination with the signal attenuation situation of the base station;

[0089] Specifically, the propagation speed of UWB signals is greatly affected by changes in air temperature and humidity. Especially in environments with high humidity, the absorption effect of the air medium on electromagnetic waves increases, resulting in a decrease in the signal propagation speed, which in turn affects the distance calculation between the base station and the tag;

[0090] In this embodiment, the temperature and humidity of the current environment are real-time collected through the deployed temperature and humidity sensors, and the change in the electromagnetic wave refractive index of the air medium is calculated based on meteorological formulas, and then the propagation rate of UWB signals in the current environment is corrected;

[0091] Furthermore, using the ranging data between base stations, calculate the signal strength attenuation curve between base stations. The signal strength attenuation curve is used to analyze the signal attenuation characteristics of the current environment. If there is a large deviation between the measured distance between base stations and the theoretical propagation distance, it indicates that the current environmental conditions have had a significant impact on signal propagation. At this time, the system needs to calculate a correction factor to compensate for this error. The calculation of the correction factor is based on historical environmental data and experimental models to ensure that the system can dynamically adjust the propagation rate of UWB signals under different environmental conditions and improve the accuracy of ranging calculations;

[0092] Specifically, first obtain the ranging data between UWB base stations, that is, the measured distance from base station A to base station B. At the same time, combined with the known coordinates of base station deployment, calculate the theoretical propagation distance from base station A to base station B, which can be calculated based on the free space propagation model. In an ideal interference-free environment, the measured distance should be close to the theoretical propagation distance. However, in the actual application environment, due to the influence of environmental factors on signal propagation, the measured distance usually deviates from the theoretical value. Therefore, by comparing the measured distance with the theoretical propagation distance, the degree of influence of the environment on signal propagation can be inferred;

[0093] Specifically, the signal strength attenuation curve between base stations is used to reflect the degree of attenuation suffered by the signal during propagation. The calculation of signal attenuation is based on the relationship between the received signal strength (RSSI, Received Signal Strength Indicator) and distance. In this method, base station A sends a UWB signal to base station B, and base station B measures and records the received signal strength. Through multiple measurements, draw a curve of signal strength changing with distance and compare it with the free space path loss model (Friis transmission equation). If the measured signal attenuation rate is much higher than the theoretical value of the free space model, it indicates that there may be wall occlusion, humidity influence or multipath interference during signal propagation;

[0094] Specifically, by storing historical environmental parameters (temperature, humidity, obstacle type, etc.) and signal propagation data, establish ranging error models in various environments. For example, in a high-humidity environment, the signal propagation rate decreases, and the ranging error may show a systematic offset; in a multipath environment, the ranging error may show random fluctuations. Combine the real-time measured signal attenuation curve and match it with historical data and experimental models to calculate the correction factor in the current environment;

[0095] Furthermore, the calculation of the correction coefficient is not only based on historical data and experimental models, but also needs to be dynamically adjusted in combination with real-time environmental data. For example, when a large change in ambient temperature and humidity is detected, the system will preferentially refer to the correction model in a high-humidity environment; when a change in the type of obstacle on the signal propagation path is detected (such as the construction of a new partition wall), the system will update the wall attenuation correction parameters. In addition, in some special environments (such as underground parking lots), the signal may be strongly interfered by metal reflections. The system will analyze the change in the angle of arrival (AoA) and received signal strength (RSSI) of the signal to determine whether further adjustment of the correction coefficient is required.

[0096] SA2: Identify the wall type in the current environment through the signal strength change rate between base stations, angle of arrival information, and known building layout data;

[0097] Specifically, in a complex indoor environment, different types of walls (such as glass walls, wooden walls, brick walls, metal walls) have different attenuation degrees for UWB signals, which directly affect the accuracy of ranging calculations. The core objective of this step is to automatically identify the wall type on the signal propagation path and adjust the signal attenuation compensation model based on the wall material characteristics. It should be noted that since the ranging distance calculation in step SA1 calculates the relative relationship between base stations, it has a synchronous relationship with the attenuation of the wall, so there is no need to perform the operation of identifying the wall type in SA2 in advance;

[0098] In this embodiment, first, calculate the signal strength change rate between base stations, that is, the attenuation rate of the signal as the distance increases. If the attenuation rate in a certain area is much higher than the theoretical value of the free space propagation model, it means that the signal has passed through an obstacle during propagation. Secondly, use the UWB angle of arrival (AoA) measurement technology, combine the data of multiple base stations, calculate the incident angle and reflection angle of the signal, and judge whether the signal has passed through the wall by comparing the known building structure layout information. Finally, through the established wall type database, match the wall type in the current environment and determine its typical attenuation value for UWB signals in combination with experimental data. Through the above method, the system can automatically judge whether the signal is affected by the wall and accurately identify the wall type, providing key input data for subsequent signal attenuation compensation.

[0099] SA3: Determine the signal attenuation characteristics according to the wall type and calculate the correction coefficient of signal attenuation;

[0100] Specifically, once the wall type on the signal propagation path is identified, the system will calculate the correction coefficient of signal attenuation based on the electromagnetic characteristics of the wall material. Different materials of walls have different absorption and reflection effects on UWB signals. For example:

[0101] Glass wall: The UWB signal has good penetrability and small attenuation;

[0102] Wooden wall: The attenuation of the signal is medium, and the attenuation degree is related to the wall thickness;

[0103] Brick wall: The signal attenuation is large. Especially when the wall thickness exceeds 20 cm, large errors may occur in the UWB signal;

[0104] Metal wall: Almost completely reflects the UWB signal, resulting in an increase in non-line-of-sight (NLOS) ranging error.

[0105] Furthermore, to calculate the correction coefficient, the system will establish a mathematical model of the wall material and signal attenuation based on experimental data, and extract the corresponding correction coefficient from the database according to the currently identified wall type. For example, if the system detects that the signal passes through a brick wall and the wall thickness is 30 cm, according to the experimental data, the system may apply an attenuation compensation factor to adjust the ranging error on this signal path.

[0106] Preferably, to further optimize the correction coefficient, the system will also combine historical ranging data and environmental change trends, and use machine learning methods to continuously update the correction model to adapt to the characteristic changes of different building environments.

[0107] S2.3: Sort the base stations according to the timestamp size , where represents the set of base station timestamps sorted from smallest to largest, represents the sorting function, and determines the sorting index of base stations in different dimensions according to the dimension type identification information of the base stations, and outputs available base stations;

[0108] Specifically, after excluding abnormal base stations, sort the timestamp data of the remaining base stations. According to the size of the timestamp, the smaller the timestamp value, the closer the base station is to the tag. Through sorting, the base station closest to the tag can be intuitively determined, and it provides a priority reference for subsequent dimension selection and positioning calculation.

[0109] Furthermore, on the basis of sorting, combined with the dimension type identification information of the base stations (such as the type identification of two-dimensional base stations, one-dimensional base stations, and zero-dimensional base stations), further classify and index base stations in different dimensions. It can effectively meet the multi-dimensional positioning requirements in complex environments. For example, in multi-dimensional intersection areas such as shopping malls, elevators, and stairs, the most suitable base stations are preferentially selected for positioning calculation.

[0110] Exemplarily, when the target tag moves to the escalator, the two-dimensional plane base stations can be quickly identified through sorting, and at the same time, the height information of the zero-dimensional base stations is referred to ensure seamless conversion between the two-dimensional plane position and the height dimension. In addition, the classification index of the base stations can also optimize the calculation amount in multi-dimensional fusion calculation, only select key base stations to participate in the calculation, avoid introducing interference from invalid data, and improve the overall positioning efficiency.

[0111] Please refer to Figure 3 , the flowchart of the position calculation method based on the position of the previous epoch in the embodiments of the present invention. In indoor UWB positioning, it will be affected by non-line-of-sight, including indoor walls and equipment, etc. In such an environment, if all base stations participate in the position calculation, large errors are bound to occur. At the same time, the indoor environment may include different dimensional regions, and the position calculation in the transition regions of different dimensions is also a difficult point;

[0112] Specifically, first, obtain the timestamp sorting index, dimension identifier, and position information of the previous epoch of the base stations. The system enters the judgment process of two-dimensional base stations: if the number of two-dimensional base stations is sufficient (at least three base stations) and the index of the nearest base station is 0, then calculate the two-dimensional position coordinates of the target using the two-dimensional base station data; if the conditions are not met, the system skips the two-dimensional calculation and enters the one-dimensional base station judgment process. In the one-dimensional positioning stage, the system detects whether the number of one-dimensional base stations meets the positioning requirements (at least two base stations) and the index of the first base station is 0. If the conditions are met, calculate the one-dimensional position coordinates of the target through the one-dimensional base stations; otherwise, enter the zero-dimensional base station calculation process. For zero-dimensional positioning, the system determines whether there are available zero-dimensional base stations and the index of the first base station is 0. If the conditions are met, directly select the position of the nearest zero-dimensional base station as the zero-dimensional position of the target. Subsequently, the system adopts different positioning result optimization strategies according to whether it is the first epoch: if it is the first epoch, take the average value of the two-dimensional, one-dimensional, and zero-dimensional positions as the final positioning result of the target; if it is not the first epoch, combine the position information of the previous epoch, calculate the Euclidean distance between the positions of different dimensions and the position of the previous epoch, and select the dimension result with the smallest distance as the final positioning result. This process effectively solves the problem of unstable positioning in the intersection area of multi-dimensional environments by preferentially calculating two-dimensional and one-dimensional positions and then zero-dimensional positions, dynamically selecting dimensions to participate in positioning, and optimizing in combination with the data of the previous epoch, while taking into account the adaptability and robustness in complex scenarios to ensure high-precision indoor positioning effects.

[0113] The specific steps of S3 are as follows:

[0114] S3.1: According to the timestamp sorting index of the available base stations, select combinations of multiple two-dimensional base stations, and calculate multiple groups of two-dimensional position coordinates according to the three-base station combination method;

[0115] In this embodiment, in order to reduce the non-line-of-sight error, during the sorting process, the base stations with higher signal strength are selected based on the timestamp value, while avoiding calculation errors introduced by low-quality base stations. The design of multi-base station combination can significantly improve the robustness of the positioning result, because even if the signals of some base stations are interfered, reliable coordinate results can still be obtained through other combinations. In addition, by restricting the number of combined base stations (for example, at most 5 base stations are selected), the calculation complexity can be effectively reduced to ensure the real-time requirements of the system.

[0116] Specifically, at least three base stations are required for two-dimensional position calculation. In order to reduce the non-line-of-sight error and the calculation amount, at most the first 5 base stations are selected to calculate multiple groups of two-dimensional coordinates through the combination of three base stations, that is, the number of combinations is the value of permutation and combination, where M is the number of two-dimensional base stations;

[0117] Furthermore, the core principle of the three-base station combination method is to utilize the geometric relationship between three base stations to solve the exact position of the target tag in the two-dimensional plane through triangulation. This method requires at least three base stations. Therefore, in actual operation, at most 5 base stations are preferentially selected from the base stations with earlier sorted timestamps for permutation and combination, so as to generate multiple groups of two-dimensional base station combinations. For each base station combination, its corresponding two-dimensional coordinates are calculated to form a two-dimensional coordinate candidate set;

[0118] Let the plane coordinates of the three base stations be 、 and , in order to establish a mathematical model, first construct the coordinate transformation matrix A between the base stations:

[0119] ;

[0120] The coordinate transformation matrix A between the base stations reflects the relative position relationship between the base stations, ensuring that the coordinate reference is consistent in the calculation. Then calculate the variables and to represent the square difference of the distances between the base stations:

[0121] ;

[0122] ;

[0123] Obtain the ranging distances between the target tag and the three base stations. Based on the ranging calculation results, construct the K matrix:

[0124] ;

[0125] Among them, represents the ranging distance between the target tag and the first base station, Indicates the distance difference between the target tag and the second base station, which is obtained by multiplying the difference between the second timestamp and the first timestamp among the three base stations by the speed of light. Indicates the distance difference between the target tag and the third base station, which is obtained by multiplying the difference between the third timestamp and the first timestamp among the three base stations by the speed of light;

[0126] Through matrix solution, the two-dimensional coordinates of the target tag Can be obtained from the following formula:

[0127] ;

[0128] S3.2: According to the previous epoch position information, eliminate the solution results in the multiple sets of two-dimensional position coordinates whose distances from the previous epoch position exceed the preset distance threshold, and take the mean value of the remaining two-dimensional position coordinates as the two-dimensional position coordinates of the current epoch;

[0129] In this embodiment, eliminating outliers can significantly reduce the positioning deviation problem caused by non-line-of-sight errors, signal reflection, or environmental interference. In addition, combined with the dynamic screening strategy of the current position information, it is possible to ensure the robustness and continuity of the system while maintaining the accuracy of the positioning result.

[0130] Specifically, by calculating the distance between each two-dimensional coordinate and the previous epoch position, the coordinate points whose distances exceed the preset threshold (for example, 3 to 5 meters) are eliminated. The setting of the distance threshold is based on the movement characteristics of the target. For example, in a crowded shopping mall, the moving speed of consumers is usually between 1 and 2 meters per second, so the displacement between adjacent epochs should not exceed the preset range.

[0131] Furthermore, after eliminating outliers, the mean value of the remaining candidate points is calculated as the two-dimensional position of the current epoch. Mean value calculation can not only smooth the signal fluctuations but also further filter out possible small-range deviation points, making the final positioning result more stable. During the screening process, the system will also dynamically adjust the threshold range according to the time interval between epochs. For example, when the target is stationary for a long time, the threshold can be appropriately reduced to enhance the accuracy of positioning.

[0132] S3.3: According to the dimension type identifier of the available base stations, select a combination of multiple one-dimensional base stations, and select the first valid solution result as the one-dimensional position coordinates of the current epoch according to the two-base-station combination method;

[0133] In this embodiment, one-dimensional positioning, as a supplement to two-dimensional positioning, can provide efficient positioning results in special scenarios and effectively reduce the positioning interruption problem caused by multi-dimensional environment conversion;

[0134] Specifically, at least two base stations are required for one-dimensional position calculation. To reduce non-line-of-sight errors, the coordinates of up to the first four base stations are used to calculate the one-dimensional coordinates through combinations of two base stations, and the combination with the first position solution is taken as the final one-dimensional position result.

[0135] Furthermore, the applicable scenarios for one-dimensional positioning include linear spaces such as corridors, escalators, and stairs. In these scenarios, the movement trajectory of the target usually unfolds along a fixed linear path, so one-dimensional positioning can quickly determine the position of the target. For example, in the scenario of a shopping mall escalator, the label signal of a consumer may be received by base stations on the upper and lower floors, and one-dimensional positioning can accurately determine the position of the consumer through the distance difference between the upper and lower base stations.

[0136] S3.4: Select the coordinates of the zero-dimensional base station with the smallest timestamp as the zero-dimensional position coordinates of the current epoch.

[0137] In this embodiment, the simplicity and reliability of zero-dimensional positioning provide a fault tolerance mechanism for the system, ensuring that in complex scenarios, even if the calculation of some dimensions fails, the system can still provide the minimum position information.

[0138] Specifically, the core of zero-dimensional positioning is to use the position of the base station with the smallest timestamp as the zero-dimensional coordinates of the target. The smallest timestamp indicates the shortest signal arrival time, usually meaning that the target is closest to this base station. Zero-dimensional positioning is mainly used to determine the existence position of the target and is applicable to position determination in narrow areas or specific scenarios.

[0139] Furthermore, typical applications of zero-dimensional positioning include elevators, shelf areas, or enclosed spaces. Through this strategy, even if the accurate two-dimensional or one-dimensional coordinates of the target cannot be calculated, the system can still quickly lock the approximate position of the target, providing support for subsequent multi-dimensional fusion.

[0140] S3.5: Calculate the label position of the current epoch based on the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates.

[0141] Specifically, through the above process, the label positions of different dimensions are calculated. If the current epoch is the first epoch, the average value of the position coordinates of each dimension is used as the final label position; otherwise, the coordinate closest to the position of the previous epoch is selected as the final coordinate position.

[0142] Furthermore, the introduction of the multi-dimensional fusion strategy solves the continuity problem in cross-dimensional positioning. For example, when a consumer enters an escalator on the first floor of a shopping mall and goes up to the second floor, the system can smoothly transition to the new dimension coordinates by fusing two-dimensional and one-dimensional positions, avoiding jumps or interruptions.

[0143] The specific steps of S3.5 are as follows:

[0144] S3.5.1: If the current epoch is the first epoch, the mean values of the two-dimensional, one-dimensional, and zero-dimensional position coordinates are used as the label position of the current epoch.

[0145] In this step, since there is no reference information from the previous epoch during the positioning process of the first epoch, it is impossible to optimize the current position information through trajectory prediction or historical data. To ensure the stability and accuracy of the initial positioning, a multi-dimensional position mean calculation method is adopted.

[0146] Specifically, after the system completes the two-dimensional, one-dimensional, and zero-dimensional positioning calculations, it will obtain position information in three different dimensions. However, since there may be certain errors or deviations in the positioning data of each dimension, simply selecting the calculation result of one dimension may lead to a large positioning error. To enhance the stability of the data and improve the initial positioning accuracy, the mean values of the two-dimensional, one-dimensional, and zero-dimensional coordinates are calculated to determine the final position of the first epoch. The main advantage of using the mean calculation is that it can effectively reduce the influence of data errors in individual dimensions and improve the reliability of the positioning. For example, in a shopping mall environment, if the user's UWB tag is in the elevator or staircase area, the traditional two-dimensional positioning method may deviate due to the signal multi-path effect. By combining the mean values of different dimensions, the true position of the user can be better reflected.

[0147] Furthermore, the multi-dimensional mean calculation method can also reduce the error accumulation when a single base station or the data of a certain dimension is abnormal, ensuring that the positioning result is smoother and will not cause too large an initial positioning deviation due to abnormal data in a certain dimension. Further, to optimize the accuracy of the mean calculation, weights can be set for the two-dimensional, one-dimensional, and zero-dimensional data. For example, if the zero-dimensional base station data is less or the signal quality is poor, the weight of the zero-dimensional data can be reduced, and the contribution of the two-dimensional or one-dimensional data can be increased to ensure the rationality and accuracy of the mean calculation.

[0148] S3.5.2: If the current epoch is not the first epoch, calculate the Euclidean distance between the position coordinates of each dimension and the position of the previous epoch, and select the dimension position with the closest Euclidean distance as the label position of the current epoch.

[0149] In this embodiment, in the case of non-first epoch, since there is already the position information of the previous epoch, the positioning accuracy can be improved through the optimization of the historical trajectory, rather than simply relying on the calculation data of the current epoch.

[0150] Specifically, in practical applications, the position change of the target object is continuous and will not experience a drastic jump in a short period of time. Therefore, the position of the previous epoch has important reference value for the current positioning. By calculating the Euclidean distances between the current two-dimensional, one-dimensional, and zero-dimensional coordinates and the position of the previous epoch, and selecting the dimension position with the smallest distance as the final label position of the current epoch, the positioning result that best matches the motion trajectory can be effectively filtered out, thereby reducing the position drift caused by data errors in individual dimensions. For example, in a factory environment, if an AGV (Automated Guided Vehicle) is moving between multiple floors, it may receive UWB signals in multiple dimensions simultaneously. At this time, directly selecting the data of a certain dimension may result in discontinuous trajectories, while by judging the Euclidean distance, it can be ensured that the selected coordinate points are consistent with the motion trajectory of the previous epoch, guaranteeing the smoothness of positioning;

[0151] Furthermore, to further improve the positioning accuracy, an adaptive weight algorithm can be introduced based on the Euclidean distance judgment. That is, in certain specific environments, such as elevators, stairs, or narrow passages, higher weights are assigned to the data of certain dimensions to enhance the adaptability of the positioning system to special scenarios. At the same time, to avoid the excessive influence of historical data on the current positioning, a dynamic time window can be set. For example, only the data of the nearest N epochs are used for reference to ensure that the system can still maintain good accuracy and real-time performance during long-term operation.

[0152] Preferably, in order to improve the stability and accuracy of the UWB positioning system, a machine learning (ML) model can also be applied. By combining historical positioning data with real-time environmental parameters, the optimal combination of three base stations (for two-dimensional positioning) and two base stations (for one-dimensional positioning) can be intelligently selected to replace the permutation and combination method. This method can dynamically adapt to environmental changes and improve the stability of the positioning system in complex scenarios (such as non-line-of-sight environments, signal interference, high-density base station deployments).

[0153] Specifically, by training a base station selection model, this model can dynamically select the optimal combination of base stations based on information such as base station signal quality, historical errors, and environmental characteristics, ensuring the highest accuracy of the calculated label position. This method includes the following steps:

[0154] Feature extraction: Construct a base station evaluation data set;

[0155] Specifically, a base station evaluation dataset needs to be constructed to ensure that the model can accurately learn the optimal strategy for base station selection. In the data collection phase, each base station is assigned feature information in multiple dimensions, which is used to describe the signal quality, reliability of the base station, and its performance under different environmental conditions. Among them, the signal quality of the base station can be quantified by RSSI (Received Signal Strength Indicator) or SNR (Signal-to-Noise Ratio), and these metrics can directly reflect the stability of the UWB signal received by the base station;

[0156] Furthermore, the system will also store the historical error data of each base station, that is, the mean and variance of the positioning error of the base station in the past epochs, to measure its long-term positioning reliability. In complex environments, such as areas with strong wall occlusion and metal reflection interference, some base stations may exhibit large errors for a long time, so it is necessary to reduce the probability of their being selected. Environmental features are also a key factor. For example, by fusing data from lidar, WiFi signals, or other sensors, the system can identify whether a base station is in a non-line-of-sight (NLOS) environment. If a base station is in an occluded state for a long time, the system will dynamically adjust its weight to reduce its impact on the final positioning calculation. All these data will be stored in the base station evaluation dataset and used as the basis for training data.

[0157] Train a machine learning model: learn the optimal base station combination rule;

[0158] Specifically, after the data collection is completed, the system needs to train a machine learning model to learn the optimal base station combination rule. The goal of training is to enable the model to identify from the past positioning data which base station combinations are most likely to bring the most accurate positioning results. When selecting a specific model, methods such as decision trees, random forests, or neural networks can be used. The decision tree model can quickly learn the impact of different base station features on positioning accuracy and generate a set of base station selection rules, such as "if the signal strength of the base station is below a certain threshold, then eliminate this base station" "if the historical error variance of the base station exceeds a certain set value, then reduce the priority of this base station", etc. The random forest model can further enhance the robustness of the model, enabling it to perform well under different environmental conditions. For more complex scenarios, the neural network model can learn the geometric relationship between base stations through multi-layer perceptron (MLP) or graph neural network (GNN) to directly predict the optimal base station combination without relying on manually set rules. During the model training process, the system will use the historical dataset for supervised learning and continuously optimize the base station selection strategy to ensure that the selected base station combination can maximize the positioning accuracy.

[0159] Real-time inference: Dynamically select the optimal base station during actual deployment;

[0160] Specifically, after the model training is completed, the system enters the real-time inference stage, that is, the optimal base station is dynamically selected during actual deployment. At each new epoch, the system collects the real-time feature information of all available base stations, including signal strength, timestamp stability, historical error, and environmental impact, etc., and inputs this data into the trained machine learning model. The model will return the optimal combination of base stations. For example, assume that in a shopping mall environment, the system detects that the signals of some base stations are affected by temporary obstacles (such as crowds, metal shelves), resulting in a decrease in RSSI and an increase in signal delay fluctuations. Traditional base station selection methods may still select these base stations according to preset rules, thus causing positioning errors. However, in this method, the machine learning model can identify the downward trend of the signal quality of these base stations and dynamically adjust the selection strategy, preferentially using base stations with stable signals for position calculation. This adaptive adjustment ability enables the system to maintain high-precision positioning ability in complex dynamic environments, avoiding the error problems caused by short-term signal fluctuations that traditional methods may encounter;

[0161] Furthermore, this model also has the ability of anomaly detection and adaptive optimization. During operation, the system continuously monitors the positioning error of the selected base stations and compares the actually calculated tag positions with the expected trajectories. If it is found that a certain base station frequently provides incorrect data within a certain period of time, the machine learning model can update its weight and reduce the priority of this base station in the subsequent base station selection process. This process can be achieved through online learning technology, enabling the system to continuously optimize the base station selection strategy during long-term operation. In addition, in the face of the situation where a base station is damaged or some base stations are offline, traditional methods may cause the system to crash due to the lack of predefined optional base stations, while this application can automatically adapt to the dynamic changes of base stations and update the base station selection strategy in real time, making the system more fault-tolerant;

[0162] Preferably, the machine learning method of this application is not only applicable to UWB positioning systems, but also can be applied to other wireless signal-based positioning technologies, such as WiFi, Bluetooth AoA, millimeter wave, etc. In intelligent factories, this method can be used to optimize the navigation system of AGVs (Automated Guided Vehicles) to ensure that the best base station is always selected for positioning in a dynamic environment. In intelligent warehousing, it can be used to accurately track the positions of goods. Even in an environment with dense shelves and strong signal interference, high-precision positioning can be maintained through an intelligent base station selection strategy. In intelligent medical scenarios, this method can be used for real-time tracking of patients and medical equipment to improve hospital management efficiency. In addition, in scenarios where GPS signals cannot be covered, such as underground parking lots, this method can also improve the positioning accuracy through dynamic base station selection, avoiding error problems caused by signal reflection and shielding;

[0163] The specific steps of S4 are as follows:

[0164] S4.1: Calculate the position information of the previous epoch and multiple epochs before it to determine the movement trajectory of the UWB tag;

[0165] In this step, to ensure the accuracy of the movement trajectory, it is first necessary to collect the UWB tag position information of the previous epoch and multiple epochs before it. Since UWB signals may be affected by environmental factors such as metal reflections and obstacle occlusions, resulting in abnormal position information for some epochs, it is necessary to use trajectory smoothing algorithms (such as sliding window filtering, weighted mean filtering, etc.) to eliminate mutation outliers. At the same time, to avoid the impact of data errors in a single dimension on the overall trajectory, this step performs multi-dimensional joint analysis, that is, not only considering the changes in planar coordinates, but also combining height information (if there is zero-dimensional data) for trajectory optimization.

[0166] Furthermore, this step introduces a historical trajectory backtracking strategy, that is, if there are large deviations in the UWB measurement values of some epochs, earlier trajectory data can be used for correction to ensure the rationality of the overall trajectory. Finally, the system constructs a trajectory curve based on the current position point and analyzes the movement trend to provide a basis for subsequent speed prediction.

[0167] S4.2: According to the movement trajectory, calculate the movement speed of the UWB tag and predict the simulated speed of the current epoch based on the time interval;

[0168] In this step, the system calculates the movement speed of the target based on the position information of the previous epoch and multiple epochs. Specifically, first calculate the displacement increment of the target between adjacent epochs based on the trajectory data, and then calculate the speed in combination with the time interval. Since the movement of the UWB tag may be in an accelerating or decelerating state, this step not only calculates the instantaneous speed, but also calculates the acceleration change rate to improve the ability to identify the movement trend.

[0169] Furthermore, the system uses the weighted time difference method, that is, performs weighted calculation on the speed data of multiple epochs to ensure that the speed estimation is more in line with the actual movement situation. In addition, when there is zero-dimensional data (such as in an elevator scenario), this step will preferentially use the height change rate of the zero-dimensional data to identify whether the target is in a floor transition state. Finally, through a time series prediction model (such as the autoregressive integrated moving average ARIMA model), the system can predict the simulated speed of the current epoch and provide accurate input for the state transition model.

[0170] S4.3: According to the simulated speed, calculate the current position prediction information through the state transition model;

[0171] In this embodiment, state transition modeling is performed based on a filtering algorithm to provide a prediction of the target's current position;

[0172] Specifically, first construct a state vector, including parameters such as the current position, velocity, and acceleration of the target, and adjust the state transition matrix according to the simulated velocity to make it more conform to the actual motion situation. In traditional UWB positioning, due to environmental interference, the ranging error is relatively large. Therefore, simply relying on current measurement data may lead to unstable position estimation. However, in this step, by fusing historical prediction information, the noise interference is effectively reduced. Especially in non-line-of-sight (NLOS) environments, such as complex scenarios like underground parking lots and storage shelves, the UWB signal may have increased errors due to multipath effects. At this time, the state transition model can be used as a compensation mechanism to make the positioning result more stable;

[0173] Furthermore, during the prediction process, the process noise covariance will be dynamically adjusted. That is, when the detected motion trajectory is stable, the weight of the process noise is reduced to improve the prediction accuracy; while when the target is in an accelerating state or sudden motion (such as turning), the weight of the process noise is increased to make the model more sensitive to motion changes.

[0174] The relevant formulas for prediction are as follows:

[0175] ;

[0176] ;

[0177] G ;

[0178] ;

[0179] Among them, represents the state transition matrix, represents the transpose matrix of the state transition matrix, represents the position and velocity of the previous epoch, represents the predicted position and velocity of the current epoch, where the initial velocity is obtained by dividing the position difference between the first epoch and the second epoch by the time difference and the initial position is the position of the second epoch, represents the covariance matrix of, represents the covariance matrix of, initially a diagonal matrix, and the diagonal elements are the velocity and variance of the position. Q represents the random process noise, represents the acceleration variance;

[0180] Specifically, the prediction process performs position prediction based on historical position data and a dynamic model. First, the current speed and predicted position of the target are calculated from the historical position data. This process involves the calculation of the state transition matrix and covariance matrix to infer the current label position based on the past motion states. First, the current position of the target is predicted according to the state transition matrix. The state transition matrix is used to represent the motion law of the target and the position change within a time interval. In this model, the predicted position of the target is closely related to the state of the previous epoch (including speed, position, etc.) and takes into account the influence of the time step. The state transition matrix linearly transforms the current position and combines the speed and position of the previous moment to obtain the predicted current position information.

[0181] Furthermore, based on the predicted historical data and measurement values for correction, the system generates an updated predicted position. This step uses the historical position data and current measurement information to adjust the prediction to ensure that the predicted position is as accurate as possible. In this process, considering the errors and covariance of the historical data, the historical data and predicted data are weighted and fused in a weighted manner. The covariance matrix represents the reliability and uncertainty of the prediction. Through these predictions and corrections, the position and speed of the target can be continuously adjusted to make the final prediction result closer to the actual trajectory.

[0182] S4.4: Correct the label position of the current epoch according to the current position prediction information to obtain the updated position of the current epoch;

[0183] Specifically, this step first calculates the Euclidean distance error between the measured position and the predicted position and sets a dynamic threshold. If the error is within the threshold range, a higher weight is assigned to the UWB measured position; otherwise, the weight of the historical predicted position is increased. For further optimization, this step introduces an adaptive filtering gain adjustment strategy, that is, when the current position prediction error is small, the confidence in the predicted position is increased; while when the historical prediction error is large, more preference is given to the current UWB measurement value. In addition, in some special scenarios, such as when the target is stationary for a short time or suddenly changes direction, this step will temporarily reduce the influence weight of the historical prediction to ensure that the positioning system can quickly adapt to changes in the motion state. By this method, the jumping point phenomenon of UWB positioning can be effectively reduced, making the positioning result smoother and more in line with the actual motion law.

[0184] The specific steps of S4.4 are as follows:

[0185] S4.4.1: Calculate the adaptive speed robust filtering gain according to the current position prediction information and the label position of the current epoch;

[0186] Specifically, when the error of the current position prediction is small, the system will increase the confidence in the predicted position, that is, assign a higher weight to the predicted position. In this way, the system can continuously track the movement trajectory of the target, smooth the historical data, and reduce the impact of short-term fluctuations on the positioning result. On the other hand, when the historical prediction error is large, it indicates that the historical data may be affected by abnormal environmental factors (such as signal occlusion, non-line-of-sight propagation, etc.). The system will give priority to adopting the current UWB measurement value, thereby improving the real-time positioning accuracy.

[0187] Furthermore, in order to handle sudden positioning changes that may occur in special scenarios, such as the target's short-term stillness or sudden change of direction, this step will temporarily reduce the influence weight of historical prediction on the current position. This mechanism ensures that when the target state changes sharply, the positioning system can quickly adapt and correct the trajectory, avoiding positioning errors caused by the inertia of historical prediction. By adaptively adjusting the weight between prediction and measurement values, the system can smoothly transition the positioning trajectory, avoiding the common jump point phenomenon in traditional UWB positioning, and making the movement of the target more in line with the actual movement law.

[0188] S4.4.2: Calculate the updated position of the current epoch according to the adaptive speed robust filtering gain;

[0189] Specifically, the calculated filtering gain value is used to weight the historical predicted position and the current UWB measurement value. Combining the ratio of the predicted position and the measured position, the updated position of the current epoch is obtained. By adjusting the filtering gain, the system can smoothly transition between dimensions, not only optimizing the positioning accuracy, but also enhancing the real-time performance and adaptability of the system. Especially in complex environments or when the motion state changes, it can quickly adjust the positioning strategy.

[0190] Embodiment 2

[0191] Please refer to Figure 4 , the present invention provides an embodiment: a UWB multi-dimensional fusion positioning system, the system includes a data acquisition module, a base station screening module, a position calculation module and a position fusion module;

[0192] The data acquisition module is used to obtain the timestamp data, base station coordinates and base station dimension type identification information from the UWB tag to each base station, and store the relevant data;

[0193] The base station screening module is used to process the data obtained by the data acquisition module, eliminate abnormal base stations based on the Z-count normalization method, sort the base stations according to the timestamp size, and determine the sorting index of different dimension base stations according to the dimension type identification information of the base stations, and output available base stations;

[0194] The position calculation module calculates the current epoch tag position based on the data of the available base stations;

[0195] The position fusion module calculates the motion trajectory of the UWB tag based on the position information of the previous epoch and multiple preceding epochs, calculates the current position prediction information using the state transition model, and fuses the tag position of the current epoch and the current position prediction information;

[0196] The position calculation module includes:

[0197] The two-dimensional position calculation unit is used to select combinations of multiple two-dimensional base stations, calculate multiple sets of two-dimensional position coordinates according to the three-base-station combination method, and use the mean value of the two-dimensional position coordinates as the two-dimensional position coordinates of the current epoch;

[0198] The one-dimensional position calculation unit is used to select combinations of multiple one-dimensional base stations, and select the first valid calculation result as the one-dimensional position coordinates of the current epoch according to the two-base-station combination method;

[0199] The zero-dimensional position calculation unit is used to select the coordinates of the zero-dimensional base station with the smallest timestamp as the zero-dimensional position coordinates of the current epoch;

[0200] The tag position calculation unit calculates the tag position of the current epoch according to the two-dimensional position coordinates, the one-dimensional position coordinates, and the zero-dimensional position coordinates.

[0201] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A UWB multi-dimensional fusion positioning method, characterized in that The UWB multi-dimensional fusion positioning method is used for indoor positioning of an object to be positioned, where a UWB tag is configured on the object to be positioned, and the UWB multi-dimensional fusion positioning method includes: Obtaining positioning data from the UWB tag of each object to be positioned to each base station and the position information of the previous epoch; Determining available base stations among all base stations according to the positioning data; Inputting the position information of the previous epoch into a preset proximity model, and the proximity model outputs the label position of the current epoch, and the label position of the current epoch is calculated according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates of the UWB tag; Fusing the current position prediction information obtained from the position information of the previous epoch with the label position of the current epoch to obtain the updated position of the current epoch.

2. The UWB multi-dimensional fusion positioning method according to claim 1, characterized in that, The positioning data includes timestamp data from the UWB tag to multiple base stations, base station coordinates, and dimension type identification information.

3. The UWB multi-dimensional fusion positioning method according to claim 2, characterized in that The determining available base stations among all base stations according to the positioning data includes: Obtaining timestamp data from the UWB tag to each base station; Calculating the normalized value of the timestamp of each base station according to the Z-count normalization method, and eliminating abnormal base stations greater than or equal to a preset standard threshold; Sorting the base stations according to the timestamp size, and determining the sorting index of base stations of different dimensions according to the dimension type identification information of the base stations, and outputting available base stations.

4. The UWB multi-dimensional fusion positioning method according to claim 3, wherein, The outputting the label position of the current epoch includes: Selecting combinations of multiple two-dimensional base stations according to the timestamp sorting index of the available base stations, and calculating multiple sets of two-dimensional position coordinates according to the three-base-station combination method; Eliminating the solution results in the multiple sets of two-dimensional position coordinates whose distance from the position of the previous epoch exceeds a preset distance threshold according to the position information of the previous epoch, and taking the mean value of the remaining two-dimensional position coordinates as the two-dimensional position coordinates of the current epoch; Selecting combinations of multiple one-dimensional base stations according to the dimension type identification of the available base stations, and selecting the first valid solution result as the one-dimensional position coordinates of the current epoch according to the two-base-station combination method; Selecting the coordinates of the zero-dimensional base station with the smallest timestamp as the zero-dimensional position coordinates of the current epoch; Calculating the label position of the current epoch according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates.

5. The UWB multi-dimensional fusion positioning method according to claim 4, wherein The calculating the label position of the current epoch according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates includes: If the current epoch is the first epoch, taking the mean value of the two-dimensional, one-dimensional, and zero-dimensional position coordinates as the label position of the current epoch; If the current epoch is not the first epoch, calculating the Euclidean distance between the position coordinates of each dimension and the position of the previous epoch, and selecting the dimension position with the closest Euclidean distance as the label position of the current epoch.

6. The UWB multi-dimensional fusion positioning method according to claim 1, wherein, The current position prediction information obtained according to the position information of the previous epoch includes: Calculating the position information of the previous epoch and multiple previous epochs, and determining the movement trajectory of the UWB tag; Calculating the movement speed of the UWB tag according to the movement trajectory, and predicting the simulated speed of the current epoch according to the time interval; Calculating the current position prediction information through a state transition model according to the simulated speed.

7. The UWB multi-dimensional fusion positioning method according to claim 6, wherein, The fusing the label position of the current epoch to obtain the updated position of the current epoch includes: Calculate the adaptive velocity robust filtering gain based on the current position prediction information and the current epoch tag position; Calculate the updated position of the current epoch according to the adaptive velocity robust filtering gain.

8. The UWB multi-dimensional fusion positioning method according to claim 4, wherein The UWB multi-dimensional fusion positioning method further includes: Obtain real-time environmental parameters and adjust the standard threshold according to the real-time environmental parameters; Optimize the contribution ratio of the base station in position calculation according to the historical signal stability and positioning accuracy of base stations in different dimensions.

9. A UWB multi-dimensional fusion positioning system for implementing a UWB multi-dimensional fusion positioning method according to any one of claims 1-8, characterized in that, The system includes a data acquisition module, a base station screening module, a position calculation module, and a position fusion module; The data acquisition module is used to obtain the timestamp data, base station coordinates, and base station dimension type identification information from the UWB tag to each base station, and store the relevant data; The base station screening module is used to process the data obtained by the data acquisition module, eliminate abnormal base stations based on the Z-count normalization method, sort the base stations according to the timestamp size, and determine the sorting index of base stations in different dimensions according to the base station dimension type identification information, and output available base stations; The position calculation module calculates the current epoch tag position based on the data of the available base stations; The position fusion module calculates the motion trajectory of the UWB tag based on the position information of the previous epoch and multiple previous epochs, calculates the current position prediction information using the state transition model, and fuses the current epoch tag position and the current position prediction information.

10. The UWB multi-dimensional fusion positioning system according to claim 9, characterized in that, The position calculation module includes: A two-dimensional position calculation unit, which is used to select combinations of multiple two-dimensional base stations, calculate multiple groups of two-dimensional position coordinates according to the three-base-station combination method, and use the mean value of the two-dimensional position coordinates as the two-dimensional position coordinates of the current epoch; A one-dimensional position calculation unit, which is used to select combinations of multiple one-dimensional base stations, and select the first valid calculation result as the one-dimensional position coordinates of the current epoch according to the two-base-station combination method; A zero-dimensional position calculation unit, which is used to select the coordinates of the zero-dimensional base station with the smallest timestamp as the zero-dimensional position coordinates of the current epoch; The tag position calculation unit calculates the current epoch tag position according to the two-dimensional position coordinates, one-dimensional position coordinates, and zero-dimensional position coordinates.

Images