An indoor shielding environment UWB and PDR fusion positioning method and system and medium

Abstract

The application provides a UWB and PDR fusion positioning method and system in an indoor shielding environment and a medium, and belongs to the technical field of indoor positioning and multi-source information fusion. The method comprises the following steps: calculating a UWB positioning track and a PDR positioning track. Taking translation, rotation, step scaling and scale change as combined parameters, a track alignment error function is constructed by using the Euclidean distance of the two tracks at each timestamp. An improved vole optimization algorithm is used to solve the optimal combined parameters. An initial population with uniform distribution is generated by mapping a Sobol difference sequence. In the iteration, the UWB track is used as the reference, the PDR track is temporarily transformed by using the combined parameters, the Euclidean distance is calculated, and the vole with the smallest distance is selected to breed the next generation. After reaching the maximum number of iterations, the PDR track is corrected by using the optimal combined parameters at one time, and the UWB track is fused and output. The application does not require a priori model, can simultaneously suppress UWB non-line-of-sight drift and PDR cumulative error, and has high positioning accuracy.

Description

Technical Field

[0001] This invention belongs to the field of indoor positioning and multi-source information fusion technology, specifically relating to a UWB and PDR fusion positioning method, system and medium in an indoor occluded environment. Background Technology

[0002] With the rapid development of the Internet of Things (IoT) and mobile internet, indoor positioning technology has wide applications in scenarios such as large shopping malls, underground parking lots, airport terminals, hospitals and nursing homes, fire and rescue operations, and smart warehousing. However, Global Navigation Satellite System (GNSS) signals are severely attenuated in indoor environments, failing to provide effective positioning services. Therefore, other technologies are needed to achieve high-precision indoor positioning. Currently, commonly used indoor positioning technologies include ultra-wideband (UWB) positioning and pedestrian path estimation (PDR) positioning.

[0003] UWB positioning calculates target location by measuring the signal flight time between the tag and the base station, offering advantages such as high accuracy and real-time performance. However, in indoor environments with obstructions, UWB signals are easily blocked by obstacles such as walls, pillars, shelves, and people, resulting in non-line-of-sight (NLOS) propagation. Simultaneously, multipath effects can interfere with ranging accuracy, leading to inflated ranging values ​​and sudden drift in positioning results. This drift is difficult to suppress effectively using conventional filtering methods, and in severe cases, can cause positioning errors of several meters for moving individuals.

[0004] PDR (Progressive Directional Positioning) utilizes acceleration and angular velocity data collected by an inertial measurement unit (IMU) to progressively advance a person's trajectory through gait detection, step length estimation, and heading angle estimation. It boasts advantages such as independence from external base stations, high autonomy, and continuous, smooth output. However, the step length estimation model has inherent errors, and gyroscope integration introduces angular drift. These errors accumulate with increasing steps, causing the trajectory to gradually deviate from the true path. After prolonged continuous movement, the cumulative error of PDR can reach tens of meters, rendering it unreliable as a standalone positioning source.

[0005] Existing UWB and PDR fusion methods mostly employ Kalman filtering or its extensions, fusing UWB positioning results as observations and PDR recursive results as state predictions. These methods suffer from the following problems: First, they rely on prior noise models and are sensitive to initial parameters. Existing methods require accurate statistical models of system and observation noise, and demand reasonable initial states, which are difficult to obtain accurately in complex real-world environments. Inappropriate initial values ​​can easily lead to the fusion results getting trapped in local optima. Furthermore, existing methods often employ point-by-point recursive filtering strategies, failing to use a unified parameter model to correct structural deviations such as translation, rotation, and scaling along the entire PDR trajectory in a single step, resulting in limited cumulative error suppression. Second, they struggle to simultaneously suppress UWB NLOS drift and PDR cumulative errors within a unified framework. UWB NLOS drift manifests as sudden jumps, while PDR cumulative errors exhibit gradual drift, making it difficult for existing methods to address both simultaneously. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides a method, system, and medium for fusion positioning of UWB and PDR in indoor occlusion environments.

[0007] To achieve the above objectives, the present invention provides a method for UWB and PDR fusion positioning in indoor obstructed environments, comprising: Acquire timestamped UWB ranging signals and IMU signals of moving targets in indoor occluded environments.

[0008] The UWB ranging signal is converted into a UWB positioning trajectory, and the IMU signal is converted into a PDR positioning trajectory. The trajectory alignment error function is constructed with the combined parameters of the PDR positioning trajectory as independent variables and the Euclidean distance between the UWB positioning trajectory points and the PDR positioning trajectory points at the same timestamp as dependent variables. The combined parameters include translation parameters, rotation parameters, step size scaling parameters, and scale change parameters.

[0009] An initial lemming population is generated using a Sobol differential sequence mapping algorithm based on an improved lemming optimization algorithm. The combined parameters represent the position vectors of the lemming population members. During the iteration process, the UWB positioning trajectory remains unchanged, and the PDR positioning trajectory is temporarily transformed using the combined parameters corresponding to the lemming population. The trajectory alignment error function value between the transformed PDR positioning trajectory and the baseline UWB positioning trajectory is calculated. The lemming with the smallest trajectory alignment error function value is selected as the optimal lemming, and the optimal lemming is used as the basis for the reproduction of the next generation population. Based on the position of the optimal lemming, the search step size of the next generation is adjusted using a preset non-uniform mutation operator. The non-uniform mutation operator decreases as the number of iterations increases.

[0010] When the maximum number of iterations is reached, the iteration stops, and the optimal combination parameters are obtained. The PDR positioning trajectory is corrected according to the optimal parameter combination. The corrected PDR positioning trajectory is combined with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment.

[0011] Preferably, the step of converting the UWB ranging signal into a UWB positioning trajectory specifically includes: The least squares support vector machine IALA-LSSVM based on the improved lemming optimization algorithm is used to identify and distinguish line-of-sight data and non-line-of-sight data in UWB ranging signals. An IALA-HKELM hybrid kernel extreme learning machine based on an improved lemming optimization algorithm was used to compensate for errors in the identified non-line-of-sight data, and the compensated non-line-of-sight distance measurement value was obtained. By using both line-of-sight (LAS) and compensated non-LAS LAS values ​​as positioning inputs, a target function for the sum of squared distance errors between the target node and multiple UWB base stations is constructed. The position coordinates of the target node are then solved using the least squares method to obtain the UWB positioning trajectory at each time step.

[0012] Preferably, the IALA-LSSVM is constructed as follows: the kernel parameters and penalty coefficients of the least squares support vector machine are encoded as position vectors of lemming individuals in the improved lemming optimization algorithm, the classification accuracy is used as the fitness function, the improved lemming optimization algorithm is used for global optimization, and the optimal parameters obtained by optimization are substituted into the least squares support vector machine; the IALA-HKELM is constructed as follows: the kernel weight coefficients, local kernel parameters, global kernel parameters and regularization coefficients of the hybrid kernel extreme learning machine are encoded as position vectors of lemming individuals in the improved lemming optimization algorithm, the error compensation accuracy is used as the fitness function, the improved lemming optimization algorithm is used for joint optimization, and the optimal parameters obtained by optimization are substituted into the hybrid kernel extreme learning machine.

[0013] Preferably, the step of converting the IMU signal into a PDR positioning trajectory specifically includes: performing peak detection based on IMU triaxial acceleration data, and completing gait detection by combining threshold filtering, detection of the existence of troughs between adjacent peaks, and time interval filtering; performing step length estimation using the Weinberg dynamic step length estimation model; obtaining the heading angle based on gyroscope integration, and correcting the heading angle through angle jump boundary detection and moving average filtering to obtain the PDR positioning trajectory.

[0014] Preferably, the step of constructing a trajectory alignment error function using the combined parameters of the PDR positioning trajectory as the independent variable and the Euclidean distance between the UWB positioning trajectory points and the PDR positioning trajectory points at the same timestamp as the dependent variable specifically includes: obtaining the corresponding trajectory points of the PDR positioning trajectory and the UWB positioning trajectory under a unified time reference; aligning the PDR positioning trajectory by translation using the difference between the starting points of the two trajectories; aligning the translated PDR trajectory by direction using a rotation matrix; determining the scale factor based on the ratio of the total length of the PDR trajectory to the UWB trajectory, and performing scale matching on the rotated PDR trajectory; and using the sum of squared Euclidean distances between corresponding points at the synchronization time as the trajectory alignment error function.

[0015] Preferably, the step size of the next generation of search is adjusted by a preset non-uniform mutation operator based on the position of the optimal lemming, and further includes: during the development phase, adjusting the search step size by a smooth development mutation operator to reduce oscillations near the optimal region.

[0016] Preferably, before converting the UWB ranging signal into a UWB positioning trajectory and the IMU signal into a PDR positioning trajectory, the method further includes: performing time synchronization and preprocessing on the UWB ranging signal and the IMU signal, mapping the two types of data to a unified time reference, and performing filtering, noise reduction, and outlier removal processing on the IMU signal.

[0017] Preferably, the step of correcting the PDR positioning trajectory according to the optimal parameter combination, and combining the corrected PDR positioning trajectory with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment, specifically includes: using the corrected PDR positioning trajectory as a continuous trajectory constraint, using the UWB positioning trajectory as an absolute position reference, and fusing to output a fusion positioning result with smoothness and absolute positioning information.

[0018] This invention also provides a UWB and PDR fusion positioning system for indoor obstructed environments, comprising: The data acquisition module is used to acquire timestamped UWB ranging signals and IMU signals of moving targets in indoor obstructed environments.

[0019] The function construction module is used to convert UWB ranging signals into UWB positioning trajectories and IMU signals into PDR positioning trajectories. Using the combined parameters of the PDR positioning trajectory as independent variables and the Euclidean distance between UWB positioning trajectory points and PDR positioning trajectory points at the same timestamp as dependent variables, a trajectory alignment error function is constructed. The combined parameters include translation parameters, rotation parameters, step size scaling parameters, and scale change parameters.

[0020] The solution module is used to generate an initial lemming population using the Sobol differential sequence mapping of an improved lemming optimization algorithm. The combined parameters represent the position vectors of the lemming population members. During the iteration process, the UWB positioning trajectory remains unchanged, and the PDR positioning trajectory is temporarily transformed by cyclically using the combined parameters corresponding to the lemming population. The trajectory alignment error function value between the transformed PDR positioning trajectory and the reference UWB positioning trajectory is calculated. The lemming with the smallest trajectory alignment error function value is selected as the optimal lemming and used as the basis for the reproduction of the next generation population. Based on the position of the optimal lemming, the search step size of the next generation is adjusted by a preset non-uniform mutation operator. The non-uniform mutation operator decreases as the number of iterations increases. When the maximum number of iterations is reached, the iteration stops, and the optimal combined parameters are obtained. The PDR positioning trajectory is corrected according to the optimal parameter combination. The corrected PDR positioning trajectory is combined with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment.

[0021] The present invention also provides a computer-readable storage medium storing a computer program, which, when loaded by a processor, is capable of executing any of the steps in the UWB and PDR fusion positioning method under indoor occlusion conditions.

[0022] The present invention provides a UWB and PDR fusion positioning method for indoor occlusion environments, which has the following advantages: The present invention first transforms the fusion positioning problem into a parameter optimization problem and uses an improved lemming optimization algorithm to solve for the optimal parameter combination. Then, it utilizes Sobol differential sequence mapping to generate a uniformly distributed initial lemming population, making the algorithm insensitive to initial parameters. During the iteration process, a non-uniform mutation operator with a step size decreasing with the number of iterations is used. Large steps are taken in the early stage to enhance global search capability and avoid getting trapped in local optima, while small steps are taken in the later stage to fine-tune and improve convergence accuracy. By using the UWB positioning trajectory as a reference and minimizing the Euclidean distance as the objective, the PDR positioning trajectory is temporarily transformed using the combination parameters corresponding to the lemming population, and the error is calculated. The optimal lemming is selected to breed the next generation, and finally, the optimal parameter combination is obtained to correct the entire PDR trajectory in one go. This method does not rely on prior statistical models of system noise and observation noise, is insensitive to initial parameters, and can eliminate the structural cumulative error of PDR as a whole through translation, rotation, step scaling, and scale changes. At the same time, it automatically suppresses the influence of isolated NLOS drift points in UWB trajectories with the goal of minimizing the overall error. Under a unified framework, it achieves simultaneous suppression of UWB non-line-of-sight drift and PDR cumulative error. It has the advantages of not requiring prior models, strong global search capability, high convergence stability, and smooth and accurate fused trajectory, which significantly improves the positioning accuracy and robustness of moving figures in indoor occluded environments. Attached Figure Description

[0023] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating a UWB and PDR fusion positioning method under indoor obstructed conditions, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the UWB and PDR fusion localization method based on the improved lemming optimization algorithm according to an embodiment of the present invention. Figure 3 This is a trajectory comparison diagram of different positioning methods in a corridor area according to an embodiment of the present invention; Figure 4 This is a cumulative error distribution diagram of different positioning methods according to embodiments of the present invention; Figure 5 This is a flowchart of trajectory alignment error optimization based on the improved lemming optimization algorithm according to an embodiment of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0026] like Figure 2 As shown, the overall framework of the fusion positioning system of this invention includes two main parts: UWB positioning trajectory calculation and PDR positioning trajectory calculation. In the UWB positioning trajectory calculation process, firstly, a least squares support vector machine based on the improved lemming optimization algorithm (IALA-LSSVM) is used to identify line-of-sight and non-line-of-sight (NLOS) signals from the UWB ranging signal. Then, a hybrid kernel extreme learning machine based on the improved lemming optimization algorithm (IALA-HKELM) is used to compensate for the identified NLOS ranging errors. Finally, the UWB positioning trajectory is calculated using the least squares method. In the PDR positioning trajectory calculation process, gait detection and step size estimation are performed based on IMU three-axis acceleration data. The heading angle obtained from gyroscope integration and filtering are combined to generate the PDR positioning trajectory. After obtaining two separate trajectories, the PDR trajectory is translated, rotated, and scaled for alignment. The improved lemming optimization algorithm (IALA) is used to optimize the trajectory alignment error, and the final fusion positioning result is output.

[0027] This invention improves the traditional lemming optimization algorithm (IALA) in three aspects: First, it uses Sobol differential sequence mapping to generate the initial population, ensuring a uniform distribution of initial parameters in the search space and overcoming the shortcomings of traditional random initialization, which easily leads to population aggregation and sensitivity to initial values. Second, it employs a non-uniform mutation operator with a step size decreasing with the number of iterations in the early stages of iteration, using large steps in the early stages to enhance global search capabilities and avoid getting trapped in local optima, and small steps in the later stages to refine the process and improve convergence accuracy. Third, it introduces a smooth development mutation operator in the later stages of iteration, allowing the generation of new individuals to simultaneously reference the current best individual and its own historical position, reducing oscillations near the optimal region through weighted averaging and improving convergence stability. These improvements collectively enhance the algorithm's global search capability, convergence speed, and solution accuracy in trajectory alignment parameter optimization.

[0028] Based on this, the present invention provides a method for fusion positioning of UWB and PDR in indoor occlusion environments, specifically as follows: Figure 1 As shown, it includes the following steps: S1. Data Acquisition and Preprocessing First, UWB ranging data, IMU three-axis acceleration data, gyroscope data, and base station coordinate information generated by the target object within the positioning area are acquired. Since the sampling frequency and timestamps of the UWB system and the PDR system may differ, UWB trajectory timestamp sequences and PDR trajectory timestamp sequences are extracted separately, and interpolation is used to map the two types of data to a unified time reference. Simultaneously, the raw IMU data undergoes preprocessing such as filtering, noise reduction, and outlier removal to improve the stability of subsequent gait detection and heading angle estimation.

[0029] Assume the PDR trajectory is at a uniform time. The position below is represented as UWB trajectories at a unified moment The position below is represented as Because the sampling periods of the PDR system and the UWB system are different, an interpolation operator is used for time synchronization, specifically as follows: ; ; In the above formula, This represents the interpolation operator for the PDR trajectory. These represent interpolation operators for UWB trajectories, used to resample the original PDR trajectory P and the original UWB trajectory U to a unified timestamp, respectively. Down; This indicates that the PDR trajectory, after interpolation and synchronization, is at a uniform time. The following position coordinates; This indicates that the UWB trajectory is synchronized at a uniform time after interpolation. The following position coordinates; This represents the k-th timestamp under a unified time base.

[0030] Through the above processing, synchronized PDR trajectory sequences and UWB trajectory sequences are formed.

[0031] S2. Constructing IALA-LSSVM and IALA-HKELM models based on an improved lemming optimization algorithm. Traditional Least Squares Support Vector Machines (LSSVMs) typically rely on manual experience or grid search to determine kernel parameters and penalty coefficients during classification, leading to problems such as blind parameter selection, low computational efficiency, and susceptibility to local optima. To address these issues, this invention employs an improved lemming optimization algorithm (IALA) to globally optimize the kernel parameters and penalty coefficients of LSSVMs. The specific improvements are as follows:

[0032] First, the kernel parameters and penalty coefficients of LSSVM are encoded as position vectors of lemming individuals in the improved lemming optimization algorithm. Then, using the classification accuracy of LSSVM on the training set as the fitness function, a uniformly distributed initial parameter combination is generated through the iterative evolution process of the improved lemming optimization algorithm using Sobol difference sequence mapping. A non-uniform mutation operator, decreasing with the number of iterations, is then used to search for the optimal parameters globally. In each iteration, the LSSVM classification model is trained using the parameter combination represented by the current lemming individual, and its classification accuracy on the validation set is calculated. The parameter combination with the highest accuracy is selected as the current optimal solution. Through multiple iterations, the kernel parameters and penalty coefficients that optimize the classification performance of LSSVM are finally output.

[0033] Substituting the optimized parameters into LSSVM yields the least squares support vector machine based on the improved lemming optimization algorithm (IALA-LSSVM). This model can automatically determine the optimal classification parameters, avoiding the blindness of manual parameter tuning and significantly improving the accuracy of recognizing line-of-sight and non-line-of-sight categories in UWB signals.

[0034] Furthermore, traditional Extreme Learning Machines (ELMs) employ a single kernel function, which limits their ability to fit NLOS ranging errors with complex nonlinear characteristics. While Hybrid Kernel Extreme Learning Machines (HKELMs) combine the advantages of local and global kernels, their kernel weight coefficients, kernel parameters (local and global), and regularization coefficients still require manual setting, making it difficult to achieve optimal compensation. To address these issues, this invention employs an improved lemming optimization algorithm (IALA) to jointly optimize the kernel weight coefficients, kernel parameters, and regularization coefficients of HKELM. The specific improvements are as follows:

[0035] First, the kernel weight coefficients, local kernel parameters, global kernel parameters, and regularization coefficients of HKELM are collectively encoded into the position vectors of individual lemmings in the improved lemming optimization algorithm. Then, using the error compensation accuracy of HKELM on the training set (e.g., the reciprocal of the root mean square error) as the fitness function, global optimization is performed through the iterative evolutionary process of the improved lemming optimization algorithm. In the early stages of iteration, the large step size characteristic of the non-uniform mutation operator is used to extensively search the parameter space, avoiding getting trapped in local optima; in the later stages of iteration, the small step size characteristic of the smooth development mutation operator is used to finely search near the optimal region, improving convergence accuracy. In each iteration, the HKELM error compensation model is constructed using the parameter combination represented by the current lemming individual, and its compensation error on the validation set is calculated. The parameter combination with the smallest error is selected as the current optimal solution. Through multiple iterations, the parameter combination that optimizes the error compensation performance of HKELM is finally output.

[0036] Substituting the optimized parameters into HKELM yields the Hybrid Kernel Extreme Learning Machine (IALA-HKELM) based on the improved lemming optimization algorithm. This model can automatically determine the optimal kernel combination and regularization parameters, more accurately fitting the complex nonlinear mapping relationship between ranging error and signal characteristics under non-line-of-sight propagation conditions, thereby obtaining more accurate NLOS ranging error compensation values.

[0037] The common technical approach of the two improved methods is to leverage the global search and adaptive parameter optimization capabilities of the Improved Lemming Optimization Algorithm (IALA) to automatically optimize key parameters in LSSVM and HKELM, respectively, overcoming the shortcomings of traditional methods that rely on human experience and are prone to getting trapped in local optima. The Sobol difference sequence mapping in the Improved Lemming Optimization Algorithm ensures the uniformity of the initial parameter distribution, while the combined use of non-uniform mutation operators and smooth development mutation operators balances global exploration with refined local search, resulting in superior performance for the optimized IALA-LSSVM and IALA-HKELM models in both NLOS recognition and error compensation tasks.

[0038] S3, UWB trajectory calculation In a preferred embodiment, LOS and NLOS data in UWB ranging data are distinguished, and error compensation is performed on the NLOS data. This invention first performs non-line-of-sight (LOS) identification and error compensation on UWB ranging signals to improve the accuracy of UWB positioning trajectories. Specifically, this invention uses a least-squares support vector machine based on an improved lemming optimization algorithm to classify and identify UWB ranging signals. This model utilizes the improved lemming optimization algorithm to globally optimize the kernel parameters and penalty coefficients of the least-squares support vector machine, avoiding the blindness of manual parameter selection. The input features of the classifier include the amplitude, signal strength, ranging residual, and signal arrival angle of the UWB signal, and the output is the category label of the current ranging signal, i.e., line-of-sight signal or non-line-of-sight signal. After obtaining the optimized classification model through offline training, fast and accurate line-of-sight and non-line-of-sight identification of real-time acquired UWB ranging signals can be performed online.

[0039] For the identified non-line-of-sight ranging data, this invention further employs a hybrid kernel extreme learning machine based on an improved lemming optimization algorithm for error compensation. This model introduces a hybrid kernel function into the extreme learning machine framework, enabling a better fit to the complex nonlinear mapping relationship between ranging error and signal characteristics under non-line-of-sight propagation conditions. The improved lemming optimization algorithm is used to optimize the kernel weight coefficients, kernel parameters, and regularization coefficients of the hybrid kernel extreme learning machine, achieving optimal fitting performance for the error compensation model. The amplitude, signal strength, ranging value, and base station geometry of the non-line-of-sight signal are used as model inputs, and the model output is the compensated ranging error value, thus obtaining the corrected non-line-of-sight ranging value.

[0040] After the above processing, the line-of-sight (LAS) ranging value and the compensated non-LAS ranging value are used together as positioning inputs to construct an objective function for the sum of squared distance errors between the target node and multiple UWB base stations. The position coordinates of the target node are then solved using the least squares method, thus obtaining the UWB positioning trajectory at each time step. Through the above non-LAS identification and compensation mechanism, the positive ranging deviation caused by non-LAS propagation in indoor occlusion environments can be effectively suppressed, significantly improving the accuracy and stability of the UWB positioning trajectory.

[0041] The compensated NLOS ranging value and LOS ranging value are used together to calculate the target location. Assume the known base station coordinates are... The target location is The measured distance from the base station to the target is Suppose there are a total of 𝑁 UWB base stations in the system, and the coordinates of the 𝑖th base station are... The target node coordinates are The measured distance from the nth base station to the target node is Then we have:

[0042] ; Based on the deviation between the measured distance and the geometric distance of each base station, an objective function for the sum of squared errors is constructed: ; Since the objective function is a nonlinear function, the Gauss-Newton iteration method is preferred for solving it, and its position update satisfies: ; In the above formula, N represents the total number of UWB base stations; 𝑖 is the base station number, 𝑖=1,2,...,N; This represents the known coordinates of the nth UWB base station in three-dimensional space; This represents the position coordinates of the target node in three-dimensional space. This represents the measured distance from the nth base station to the target node; The objective function representing the sum of squared distance errors is the sum of squares of the differences between the measured distances and the geometric distances of all base stations. The value represents the update amount of the target node's position; H represents the Jacobian matrix corresponding to the objective function, where each element is the partial derivative of the objective function with respect to each component of the target position coordinates; r represents the residual vector, which is the difference between the measured distance of each base station and the geometric distance corresponding to the current estimated position. By iteratively updating the estimated value of the target node's position, each iteration gradually reduces the error, eventually approximating the true position coordinates of the target node, thus obtaining the UWB positioning trajectory at continuous time points.

[0043] S4, PDR trajectory calculation Gait detection, step length estimation, and heading angle estimation are performed based on IMU data. Specifically, firstly, peak detection is performed using the acceleration sequence, and spurious peaks caused by noise are filtered out by amplitude thresholding. Then, secondary filtering is achieved by detecting whether there are troughs between adjacent peaks and whether the time interval between adjacent peaks exceeds a preset threshold, thus obtaining reliable gait events. Subsequently, a Weinberg dynamic step length estimation model is used to estimate the step length based on the local acceleration maxima and minima within each gait cycle. The heading angle is obtained by gyroscope integration. When the change in heading angle between adjacent time points exceeds a preset threshold, it is determined as an angle jump and boundary correction is performed. The heading angle sequence is then smoothed using a moving average filter. The position is recursively calculated based on the step length and heading angle to obtain the PDR positioning trajectory.

[0044] In a two-dimensional planar positioning scenario, let the step size of the nth step be l. k The step size scaling factor is 𝛼, and the heading angle at the 𝑘th step is 𝜃 𝑘 Then the PDR position at time n satisfy: ; ; The heading angle is obtained by gyroscope integration and satisfies: ; In the above formula, Let represent the position coordinates of the PDR positioning trajectory on the two-dimensional plane at time k, where The x-axis component. y is the component along the y-axis; k represents the step number determined by gait detection; α represents the step size at step k, which is calculated by the Weinberg dynamic step size estimation model based on the acceleration extrema; α represents the step size scaling factor, used to adaptively scale and adjust the PDR step size. The heading angle at step k is obtained by integrating the gyroscope and then filtering and correcting it. This represents the heading angle increment corresponding to the k-th step, which is the integral value of the angular velocity of the gyroscope during the k-th detection cycle. By applying a moving average filter to the heading angle, the impact of short-term noise and sudden changes on the PDR positioning trajectory can be reduced.

[0045] S5, Trajectory Alignment and Error Modeling After obtaining the UWB trajectory and PDR trajectory respectively, the PDR trajectory is adaptively adjusted to geometrically align with the UWB trajectory under a unified time reference. First, let the starting point of the PDR trajectory be P0 and the starting point of the UWB trajectory be 𝑈0. Then, the PDR trajectory is translated and aligned using the difference in starting points. The translated PDR trajectory is as follows:

[0046] ; Further, let the rotation angle be φ, and the corresponding rotation matrix is: ; The rotated PDR trajectory is then: ; To further eliminate the scale difference between the PDR trajectory and the UWB trajectory, a scale factor α is introduced to obtain the scale-adjusted PDR trajectory: ; At the synchronization time The trajectory alignment error function is constructed using the sum of squared Euclidean distances between the UWB trajectory points and the adjusted PDR trajectory points: ; In the above formula, P k P0 represents the original position coordinates of the PDR trajectory at time k; P0 represents the starting point coordinates of the PDR trajectory; U0 represents the starting point coordinates of the UWB trajectory. This represents the coordinates of the point at time k on the PDR trajectory after translation and alignment. This indicates the rotation angle, used to correct directional deviations between the PDR and UWB trajectories. Denotes a two-dimensional rotation matrix, where and Rotation angles The cosine and sine values; represents the coordinates of the k-th point of the PDR trajectory after rotation and alignment; s represents the scale factor, used to eliminate the scale difference between the PDR trajectory and the UWB trajectory, and its value is determined by the ratio of the total length of the PDR trajectory to the total length of the UWB trajectory. This represents the coordinates of the point at time k in the scaled PDR trajectory. The trajectory alignment error function is defined as the sum of squared Euclidean distances between UWB trajectory points and adjusted PDR trajectory points. By minimizing this error function, the step scaling parameter, rotation parameter, and scale parameter can be jointly optimized, thereby improving the fusion positioning accuracy.

[0047] S6. Parameter Solving Based on the Improved Lemming Optimization Algorithm Translation parameters, rotation parameters, step scaling parameters, and scale parameters are collectively encoded into the position vector of an individual lemming.

[0048] like Figure 5 As shown, in this embodiment, Sobol differential sequence mapping is first used to generate an initial lemming population to improve the uniformity of the initial individual distribution in the search space. Then, the trajectory alignment error function is used as the fitness function to evaluate the fitness of each lemming individual. During the iteration process, exploration behavior and development behavior are executed according to the current search stage. A non-uniform mutation operator is introduced in the foraging stage and a smooth development mutation operator is introduced in the development stage to balance global search capability and local fine optimization capability. Finally, the parameter combination that minimizes the trajectory alignment error function is output as the optimal alignment parameter.

[0049] In this embodiment, a single lemming individual is encoded as: ; Where Δ𝑥 and Δ𝑦 represent translation parameters, 𝜑 represents rotation parameters, 𝛼 represents step scaling parameters, and 𝑠 represents scale parameters.

[0050] S7, Fusion Positioning Output Based on the optimal parameters obtained from the improved lemming optimization algorithm, the PDR trajectory is finally corrected, and the corrected PDR trajectory is fused with the UWB trajectory to output the final localization result of the target object. In one embodiment, the corrected PDR trajectory can be used as a continuous trajectory constraint, and the UWB trajectory can be used as an absolute position reference, thereby obtaining a fused trajectory with high smoothness and high absolute positioning accuracy.

[0051] In an experimental implementation of a corridor scenario, an IMU inertial navigation module was used to collect PDR data, and walking tests were conducted within a positioning area consisting of seven UWB base stations. Figure 3 As shown, in the experimental path in the corridor area, the positioning trajectories of standalone PDR positioning, standalone UWB positioning, extended Kalman filter fusion positioning, comparative fusion method, and the method of the present invention were compared. The fusion trajectory obtained by the method of the present invention is generally closer to the actual movement path. Correspondingly, the average positioning error can be further reduced from 1.99 meters for standalone PDR positioning, 1.25 meters for standalone UWB positioning, 0.76 meters for extended Kalman filter fusion positioning, and 0.61 meters for comparative fusion method to 0.35 meters. This indicates that by adaptively adjusting the PDR trajectory and combining it with the improved lemming optimization algorithm to optimize trajectory alignment error, the impact of PDR cumulative error and UWB non-line-of-sight positioning drift on the final positioning result can be effectively reduced.

[0052] like Figure 4 As shown, the cumulative error distribution curves of each positioning method further verify the advantages of the method of the present invention in terms of positioning accuracy. The horizontal axis represents the positioning error value, and the vertical axis represents the cumulative probability. From... Figure 4 As can be seen, the error distribution range of standalone PDR positioning is the largest, with 80% of its positioning point errors within 2.5 meters; standalone UWB positioning has approximately 80% of its positioning point errors within 1.5 meters; extended Kalman filter fusion positioning has approximately 80% of its positioning point errors within 1.0 meter; the comparative fusion method has approximately 80% of its positioning point errors within 0.8 meters; while the method of this invention has approximately 80% of its positioning point errors within 0.5 meters. Its cumulative error distribution curve is located on the far left and rises the fastest, indicating that the positioning error of the method of this invention is the smallest overall and the error distribution is the most concentrated. Compared with standalone PDR positioning, standalone UWB positioning, extended Kalman filter fusion positioning, and comparative fusion method, the average positioning accuracy of the method of this invention is improved by 82.4%, 72.0%, 53.9%, and 42.6%, respectively, fully verifying the effectiveness and superiority of the method of this invention in simultaneously suppressing PDR cumulative error and UWB non-line-of-sight positioning drift.

[0053] Compared with the prior art, the advantages of the present invention are as follows: On the one hand, by adaptively adjusting the step size of the PDR trajectory and introducing joint alignment of translation, rotation, and scale, the cumulative error of PDR can be suppressed at the trajectory level, improving the consistency between the PDR trajectory and the UWB trajectory. On the other hand, by transforming the trajectory alignment process into a multi-parameter joint optimization problem and using an improved lemming optimization algorithm for global search and local fine optimization, the dependence of traditional fusion methods on prior noise models can be reduced, improving the fusion positioning accuracy and robustness in complex environments. Furthermore, in the preferred embodiment, the NLOS error in the UWB ranging data can be compensated before position calculation, thereby further reducing the positioning drift caused by non-line-of-sight propagation and improving the stability and reliability of the fusion positioning results.

[0054] Based on the same inventive concept, this invention also provides a UWB and PDR fusion positioning system for indoor obstructed environments, comprising: The data acquisition module is used to acquire timestamped UWB ranging signals and IMU signals of moving targets in indoor obstructed environments.

[0055] The function construction module is used to convert UWB ranging signals into UWB positioning trajectories and IMU signals into PDR positioning trajectories. Using the combined parameters of the PDR positioning trajectory as independent variables and the Euclidean distance between UWB positioning trajectory points and PDR positioning trajectory points at the same timestamp as dependent variables, a trajectory alignment error function is constructed. The combined parameters include translation parameters, rotation parameters, step size scaling parameters, and scale change parameters.

[0056] The solution module is used to generate an initial lemming population using the Sobol differential sequence mapping of an improved lemming optimization algorithm. The combined parameters represent the position vectors of the lemming population members. During the iteration process, the UWB positioning trajectory remains unchanged, and the PDR positioning trajectory is temporarily transformed by cyclically using the combined parameters corresponding to the lemming population. The trajectory alignment error function value between the transformed PDR positioning trajectory and the reference UWB positioning trajectory is calculated. The lemming with the smallest trajectory alignment error function value is selected as the optimal lemming and used as the basis for the reproduction of the next generation population. Based on the position of the optimal lemming, the search step size of the next generation is adjusted by a preset non-uniform mutation operator. The non-uniform mutation operator decreases as the number of iterations increases. When the maximum number of iterations is reached, the iteration stops, and the optimal combined parameters are obtained. The PDR positioning trajectory is corrected according to the optimal parameter combination. The corrected PDR positioning trajectory is combined with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment.

[0057] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the UWB and PDR fusion positioning method provided above in an indoor occlusion environment.

[0058] Specific limitations regarding the computational system for the UWB and PDR fusion positioning method under indoor occlusion environments can be found in the above-mentioned limitations of the UWB and PDR fusion positioning method under indoor occlusion environments, and will not be repeated here. Each module in the aforementioned UWB and PDR fusion positioning system under indoor occlusion environments can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0059] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Furthermore, the above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for fusion positioning of UWB and PDR in an indoor occlusion environment, characterized in that, Includes the following steps: Acquire timestamped UWB ranging signals and IMU signals of moving targets in indoor occluded environments; The UWB ranging signal is converted into a UWB positioning trajectory, and the IMU signal is converted into a PDR positioning trajectory. The trajectory alignment error function is constructed with the combined parameters of the PDR positioning trajectory as the independent variable and the Euclidean distance between the UWB positioning trajectory point and the PDR positioning trajectory point at the same timestamp as the dependent variable. The combined parameters include translation parameters, rotation parameters, step size scaling parameters, and scale variation parameters; An initial lemming population is generated using a Sobol differential sequence mapping algorithm based on an improved lemming optimization algorithm. The combined parameters represent the position vectors of the lemming population members. During the iteration process, the UWB positioning trajectory remains unchanged, and the PDR positioning trajectory is temporarily transformed using the combined parameters corresponding to the lemming population. The trajectory alignment error function value between the transformed PDR positioning trajectory and the baseline UWB positioning trajectory is calculated. The lemming with the smallest trajectory alignment error function value is selected as the optimal lemming, and the optimal lemming is used as the basis for the reproduction of the next generation population. Based on the position of the optimal lemming, the search step size of the next generation is adjusted using a preset non-uniform mutation operator. The non-uniform mutation operator decreases as the number of iterations increases. When the maximum number of iterations is reached, the iteration stops, and the optimal combination parameters are obtained. The PDR positioning trajectory is corrected according to the optimal parameter combination. The corrected PDR positioning trajectory is combined with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment.

2. The method for fusion positioning of UWB and PDR in an indoor occlusion environment according to claim 1, characterized in that, The process of converting the UWB ranging signal into a UWB positioning trajectory specifically includes: The least squares support vector machine IALA-LSSVM based on the improved lemming optimization algorithm is used to identify and distinguish line-of-sight data and non-line-of-sight data in UWB ranging signals. An IALA-HKELM hybrid kernel extreme learning machine based on an improved lemming optimization algorithm was used to compensate for errors in the identified non-line-of-sight data, and the compensated non-line-of-sight distance measurement value was obtained. By using both line-of-sight (LAS) and compensated non-LAS LAS values ​​as positioning inputs, a target function for the sum of squared distance errors between the target node and multiple UWB base stations is constructed. The position coordinates of the target node are then solved using the least squares method to obtain the UWB positioning trajectory at each time step.

3. The UWB and PDR fusion positioning method for indoor obstructed environments according to claim 2, characterized in that, The IALA-LSSVM is constructed as follows: the kernel parameters and penalty coefficients of the least squares support vector machine are encoded as position vectors of lemming individuals in the improved lemming optimization algorithm. The classification accuracy is used as the fitness function, and the improved lemming optimization algorithm is used for global optimization. The optimal parameters obtained by optimization are then substituted into the least squares support vector machine. The IALA-HKELM is constructed as follows: the kernel weight coefficients, local kernel parameters, global kernel parameters, and regularization coefficients of the hybrid kernel extreme learning machine are encoded as position vectors of lemming individuals in the improved lemming optimization algorithm. The error compensation accuracy is used as the fitness function, and the improved lemming optimization algorithm is used for joint optimization. The optimal parameters obtained by optimization are then substituted into the hybrid kernel extreme learning machine.

4. The UWB and PDR fusion positioning method for indoor obstructed environments according to claim 1, characterized in that, The process of converting the IMU signal into a PDR positioning trajectory specifically includes: peak detection based on IMU three-axis acceleration data, combined with threshold filtering, trough existence detection between adjacent peaks, and time interval filtering to complete gait detection; step size estimation using the Weinberg dynamic step size estimation model; obtaining the heading angle based on gyroscope integration, and correcting the heading angle through angle jump boundary detection and moving average filtering to obtain the PDR positioning trajectory.

5. The UWB and PDR fusion positioning method for indoor obstructed environments according to claim 1, characterized in that, The method of constructing a trajectory alignment error function, using the combined parameters of the PDR positioning trajectory as independent variables and the Euclidean distance between UWB positioning trajectory points and PDR positioning trajectory points at the same timestamp as the dependent variable, specifically includes: obtaining the corresponding trajectory points of the PDR positioning trajectory and the UWB positioning trajectory under a unified time reference; aligning the PDR positioning trajectory by translation using the difference between the starting points of the two trajectories; aligning the translated PDR trajectory by direction using a rotation matrix; determining the scale factor based on the ratio of the total length of the PDR trajectory to the UWB trajectory, and performing scale matching on the rotated PDR trajectory; and using the sum of squared Euclidean distances between corresponding points at the synchronization time as the trajectory alignment error function.

6. The UWB and PDR fusion positioning method for indoor obstructed environments according to claim 1, characterized in that, The method of adjusting the search step size of the next generation based on the position of the optimal lemming through a preset non-uniform mutation operator also includes: during the development phase, adjusting the search step size through a smooth development mutation operator to reduce oscillations near the optimal region.

7. The method for fusion positioning of UWB and PDR in an indoor occlusion environment according to claim 1, characterized in that, Before converting the UWB ranging signal into a UWB positioning trajectory and the IMU signal into a PDR positioning trajectory, the method further includes: performing time synchronization and preprocessing on the UWB ranging signal and the IMU signal, mapping the two types of data to a unified time reference, and performing filtering, noise reduction, and outlier removal processing on the IMU signal.

8. The UWB and PDR fusion positioning method for indoor obstructed environments according to claim 1, characterized in that, The step of correcting the PDR positioning trajectory based on the optimal parameter combination, and combining the corrected PDR positioning trajectory with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment, specifically includes: using the corrected PDR positioning trajectory as a continuous trajectory constraint, using the UWB positioning trajectory as an absolute position reference, and fusing to output a fusion positioning result with smoothness and absolute positioning information.

9. A UWB and PDR fusion positioning system for indoor obstructed environments, characterized in that, include: The data acquisition module is used to acquire timestamped UWB ranging signals and inertial measurement unit (IMU) signals of moving targets in indoor obstructed environments. The function construction module is used to convert UWB ranging signals into UWB positioning trajectories and IMU signals into PDR positioning trajectories. A trajectory alignment error function is constructed using the combined parameters of the PDR positioning trajectory as independent variables and the Euclidean distance between UWB positioning trajectory points and PDR positioning trajectory points at the same timestamp as the dependent variable. The combined parameters include translation parameters, rotation parameters, step size scaling parameters, and scale variation parameters. The solution module is used to generate an initial lemming population using the Sobol differential sequence mapping of an improved lemming optimization algorithm. The combined parameters represent the position vectors of the lemming population members. During the iteration process, the UWB positioning trajectory remains unchanged, and the PDR positioning trajectory is temporarily transformed by cyclically using the combined parameters corresponding to the lemming population. The trajectory alignment error function value between the transformed PDR positioning trajectory and the reference UWB positioning trajectory is calculated. The lemming with the smallest trajectory alignment error function value is selected as the optimal lemming, and the optimal lemming is used as the basis for the reproduction of the next generation population. Based on the position of the optimal lemming, the search step size of the next generation is adjusted by a preset non-uniform mutation operator. The non-uniform mutation operator decreases as the number of iterations increases. When the maximum number of iterations is reached, the iteration stops, and the optimal combination parameters are obtained. The PDR positioning trajectory is corrected according to the optimal parameter combination. The corrected PDR positioning trajectory is combined with the UWB positioning trajectory to output the fusion positioning result of the target moving object in an indoor occluded environment.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 8.

Images