Indoor construction robot ultra-wideband non-line-of-sight positioning error judgment and mitigation method

Abstract

The application discloses an ultra-wideband non-line-of-sight positioning error judgment and mitigation method for an indoor construction robot. As an effective classification algorithm, a support vector machine can realize effective distinction of line-of-sight and non-line-of-sight data by analyzing statistical characteristic differences such as average values and standard deviations of measurement data. The mitigation method combines the square of the mean difference and the standard deviation ratio to comprehensively evaluate the difference in concentration trend and expansibility of the line-of-sight and non-line-of-sight data, in view of the significant difference in error statistical characteristics under line-of-sight and non-line-of-sight conditions. In an indoor environment seriously affected by non-line-of-sight, the system may completely lack line-of-sight data. In this case, the fluctuation of non-line-of-sight data is converted into a dimensionless factor by the error relative coefficient of variation to reflect the fluctuation of non-line-of-sight data. The application can reduce the positioning error caused by non-line-of-sight, improve the positioning accuracy and robustness, and is suitable for accurate positioning of an indoor mobile construction robot in a complex indoor environment.

Description

Technical Field

[0001] This invention relates to an ultra-wideband positioning method for indoor mobile construction robots, and more particularly to a method for judging and mitigating ultra-wideband non-line-of-sight positioning errors of indoor mobile construction robots. Background Technology

[0002] Construction robots, as a key component of intelligent construction, improve operational efficiency, construction accuracy, and operational safety through the automation and intelligence of the construction process. This necessitates precise self-localization for indoor construction robots. However, complex indoor construction environments contain various obstacles of different sizes, such as material storage areas, pillars, indoor construction equipment, and personnel movement. These obstacles significantly increase the probability of non-line-of-sight (NLOS) interference, further complicating the ultra-wideband (UWB) precise localization of indoor construction robots. Therefore, researching the assessment and mitigation of NLOS errors is a core issue that urgently needs to be addressed for the high-precision localization of indoor mobile construction robots.

[0003] Currently, research approaches for identifying and mitigating NLOS errors can be broadly categorized into two types: The first, based on UWB channel characteristics, focuses on the accurate classification and correction of NLOS signals. This method extracts key channel features by analyzing signal strength, propagation delay, and multipath effects, and then uses machine learning algorithms to determine whether the signal is affected by NLOS conditions. However, this method has several limitations, primarily requiring complex algorithms and high computational resources during signal processing, increasing system complexity and cost. The second approach, based on environmental modeling, aims to construct accurate environmental maps to reduce NLOS errors. This method uses laser scanning or photogrammetry to establish an environmental model. After collecting signal data in the actual environment, the system matches this data with the environmental model to identify and correct errors. However, this method also has limitations; the accuracy of the environmental model depends on accurate scanning of the actual environment. If the model is not refined enough or not updated in a timely manner, it may lead to positioning errors.

[0004] In view of the various shortcomings of the above-mentioned NLOS error judgment and mitigation methods, in order to achieve precise UWB positioning of mobile construction robots in indoor building environments and improve the work efficiency, construction accuracy and work safety of indoor construction, this invention with practical value has finally been created after continuous research and design, and repeated experiments and improvements. Summary of the Invention

[0005] This invention addresses the problem of improving UWB positioning accuracy for indoor mobile construction robots in complex indoor building environments by providing a method for judging and mitigating NLOS errors in UWB positioning of indoor mobile construction robots. The judgment method offers accurate feature analysis of ranging data, high recognition rate, good universality, and high computational speed. The mitigation method has broad applicability, effectively mitigating the impact of varying degrees of NLOS on measurement data, thereby improving positioning accuracy and robustness.

[0006] The principle of this invention: Due to multipath effects, UWB signals are affected by factors such as signal reflection and diffraction, leading to deviations in measurement data. Compared to line-of-sight (LOS) propagation, multipath effects are typically more pronounced and signal paths are more complex in NLOS environments, resulting in greater deviations in NLOS measurement data compared to LOS data. Support Vector Machines (SVMs), as a classification method, can distinguish between LOS and NLOS data by utilizing differences in statistical characteristics such as the mean and standard deviation of measurement data.

[0007] Given the significant differences in error statistics between LOS and NLOS conditions, this study combines the mean difference and the standard deviation ratio to comprehensively measure the differences in central tendency and expansion between LOS and NLOS data. The mean difference reflects the systematic difference in data offset between the two, while the standard deviation ratio further amplifies the impact of the difference in volatility, thus more sensitively capturing inconsistencies in expansion. In indoor environments severely affected by NLOS, the system may completely lack LOS data. In this case, the relative coefficient of variation of error is used to transform the volatility of NLOS data into a dimensionless factor to reflect the volatility of NLOS data. Based on the above analysis, an error mitigation factor is designed to alleviate the negative impact of NLOS errors on positioning accuracy.

[0008] Due to NLOS signal propagation, some base stations may not provide sufficient effective ranging data, preventing the system from simultaneously obtaining accurate positioning information from multiple base stations, thus resulting in insufficient data for positioning. A random forest (RF) model is used to predict UWB measurement data to supplement the completeness of the dataset.

[0009] This invention provides a solution for judging and mitigating non-line-of-sight (LOS) errors in UWB localization of indoor mobile construction robots in complex building environments. It utilizes an SVM classification model to calculate the mean and standard deviation of UWB measurement data within a unit sliding window, constructing feature vectors, which are then standardized using a feature scaler. The optimal combination of SVM model penalties, kernel coefficients, and kernel hyperparameters is optimized through grid search and cross-validation to complete SVM model training.

[0010] Step 1: Divide the UWB measurement simulation dataset into a 70% training set and a 30% test set, and calculate the mean and standard deviation of the training set within the sliding window K. The mean and standard deviation are as follows:

[0011]

[0012]

[0013] Among them, Mean K x is the average value within the sliding window; i SD represents the i-th data point within the sliding window; N represents the total number of data points within the sliding window; K This represents the standard deviation within the sliding window.

[0014] Step 2: The feature vectors calculated in Step 1 are standardized using the feature extractor StandardScaler. Training employs a grid search with three-fold cross-validation. The training set is divided into three subsets; one subset is selected sequentially as the validation set, and the other two subsets are used as the training set for three independent training rounds to determine the optimal hyperparameter combination. The hyperparameter search range is as follows:

[0015]

[0016] Here, C is the penalty parameter, which determines the model's tolerance for misclassification. A larger value reduces misclassification but increases the risk of overfitting, while a smaller value improves generalization ability but increases the risk of underfitting. γ is the kernel function coefficient, which controls the influence range of a single sample. A larger value captures more details but increases the risk of overfitting, while a smaller value provides smoother data but increases the risk of underfitting. Kernel is the kernel function, which determines the mapping method of the feature space. The RBF kernel is powerful and versatile, suitable for most non-linear data, with strong mapping ability but prone to overfitting, and simple parameter tuning but high computational cost. The poly kernel is more flexible, but parameter tuning is complex and may overfit higher-order polynomials. The appropriate setting of these parameters is crucial to the complexity and generalization ability of the SVM classification model.

[0017] Step 3: Construct a confusion matrix and learning curve using the test set to evaluate the model, and save the trained SVM classification model.

[0018] The training set was divided into NLOS and LOS training sets by the SVM classification model. The distance difference and time interval between different datasets within a unit sliding window after classification were calculated as additional features. Three features from the two classes of features trained in the SVM classification model were randomly selected as the basis features for the optimal split point. The optimal hyperparameter combination of the number of trees, the maximum tree depth, the minimum number of classification samples, and the minimum number of node samples was optimized through grid search and cross-validation to complete the training of RF regression models for different datasets.

[0019] Step 4: Calculate the distance difference and timestamp time interval between the NLOS and LOS training sets within the sliding window K as additional features. The distance difference and time interval are as follows:

[0020] Δx k =x k -x k-1 ,

[0021] ΔT k =T k -T k-1 ,

[0022] Where, Δx k x represents the distance difference within the sliding window; k ΔT represents the distance difference value of the k-th sliding window. k T represents the time interval between timestamps within the sliding window. k This represents the time interval value for the k-th sliding window.

[0023] Step 5: Randomly select three features as the basis features for the best segmentation point. Training uses three-fold cross-validation and grid search to determine the optimal hyperparameter combination. The hyperparameter search range is as follows:

[0024]

[0025] Where T is the number of trees; more trees can improve the stability and accuracy of the model, but training and prediction times will also increase; D max The maximum depth of the tree is N. A larger depth allows the model to capture complex features but is prone to overfitting, while a smaller depth can prevent overfitting but is prone to underfitting, reducing model performance. split N represents the minimum number of samples required for a split. Smaller values ​​allow the tree to split more times, resulting in a more complex model, but also making it more prone to overfitting. Larger values ​​reduce splits and enhance the model's generalization ability, but are more prone to underfitting. leaf The minimum number of samples for each leaf node is denoted by . A larger value makes the tree smoother and avoids overfitting, but it may overlook important details in the data. By adjusting these parameters, a trade-off can be struck between training time, model complexity, and generalization ability.

[0026] Step 6: Evaluate the model using the test set and save the trained RF regression model.

[0027] The calculation method for the error mitigation factor is determined based on the proportion of LOS data within a unit sliding window. When the proportion of LOS data is greater than zero, the ratio of the mean difference and standard deviation of LOS and NLOS measurement data is combined, and an exponential decay method is used to make the error mitigation factor more sensitive to changes in the degree of difference. At the same time, a weighting parameter λ is introduced to control the combined influence of the two on the final result. If the proportion of LOS is zero, the relative coefficient of variation of the NLOS data error is used directly, and an adjustment parameter γ is introduced to control the influence intensity of the relative coefficient of variation. Finally, the error mitigation factor is output through an exponential function.

[0028] Step 7: When the proportion of LOS measurement data is greater than 0, first calculate the mean difference Δ. μ The difference between LOS and NLOS is measured and calculated as follows:

[0029]

[0030] Where, μ LOS The mean error of the LOS measurement data; μ NLOS This represents the mean error of the NLOS measurement data.

[0031] Step 8: Calculate the standard deviation ratio Δ σ The difference in scalability between LOS and NLOS measurement data is measured using the following formula:

[0032]

[0033] Where, σ LOS σ represents the standard deviation of the LOS measurement data. NLOS This represents the standard deviation of NLOS measurement data.

[0034] Step 9: Calculate the error mitigation factor using the ratio of the mean difference to the standard deviation, as shown in the following formula:

[0035]

[0036] Where L is the error mitigation factor; λ is the control factor used to adjust the influence intensity of the error mitigation factor.

[0037] Step 10: Combining the ratio of the mean difference to the standard deviation, we can derive the following formula:

[0038]

[0039] After extensive simulation experiments and optimization, λ was set to 1.5, and the error mitigation factor was calculated to be 0.76.

[0040] Step 11: When the proportion of LOS measurement data is 0, use the relative coefficient of variation of NLOS measurement data and calculate the error mitigation factor in the absence of LOS measurement data according to the formula:

[0041]

[0042] Here, γ is an adjustment parameter that controls the intensity of the error mitigation factor's influence. After extensive experimental optimization, the value of γ was set at 2.8, from which the error mitigation factor was calculated to be 0.61.

[0043] Step 12: UWB measurement data adjusted with error mitigation factor The calculation method is as follows:

[0044]

[0045] Where, x LOS To adjust the average value of LOS measurement data within the previous sliding window K; x NLOS To adjust the average value of NLOS measurements within the previous sliding window K; η LOS η represents the proportion of LOS measurement data within the sliding window k. LOS The calculation method is as follows:

[0046]

[0047] Where, N LOS N represents the number of LOS measurement data points within the sliding window K; Total This represents the total amount of UWB measurement data within the sliding window K.

[0048] The advantages of this invention compared with the prior art are: the judgment method has accurate feature analysis of ranging data, high recognition rate, good universality and high calculation speed; the mitigation method has wide applicability and can effectively mitigate the impact of measurement data of different degrees of NLOS, thereby improving positioning accuracy and robustness. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a technical roadmap of the NLOS positioning error judgment and mitigation method in the embodiments of the present invention;

[0051] Figure 2This is a roadmap of UWB data acquisition and preprocessing technology in this invention embodiment;

[0052] Figure 3 This is a flowchart of the UWB positioning process in an embodiment of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0054] The steps of this invention are as follows: Four positioning base stations of the positioning system are deployed in the positioning area according to the specific indoor environment; a positioning tag is loaded onto the indoor mobile construction robot; the main base station collects communication data between each base station and the tag to obtain the original UWB measurement dataset; the original data is preprocessed to obtain a preprocessed UWB measurement dataset; a simulation environment is constructed based on the mapping of the on-site indoor environment to obtain a UWB simulation dataset; SVM and RF models are trained using the simulation dataset, error mitigation factors are derived, and a non-line-of-sight positioning error judgment and mitigation method is designed; the invention method is used to mitigate the UWB measurement dataset, thereby achieving accurate positioning of the indoor mobile construction robot.

[0055] like Figure 2 As shown: using base station location information (x i ,y i (i = 1, 2, 3, 4) and two-way time-of-flight ranging d TW-TOFi Distance data was obtained for (i = 1, 2, 3, 4), and the data was preprocessed using linear interpolation, quartile range method, and K-means clustering algorithm. Where d TW-TOFi This represents the distance measured between the i-th base station and the positioning tag.

[0056] Step 1: Measure the distance between each base station and the positioning tag using two-way time-of-flight measurement. The distance is shown in the following formula:

[0057]

[0058] Where c is the speed of signal propagation, which is the speed of light, approximately 3 × 10⁻⁶. 8 m / s; T roundA T is the round-trip time from the base station to the tag. RB T is the one-way time from the tag to the base station; roundB T is the round-trip time from the tag to the base station. RA This is the one-way time from the base station to the tag.

[0059] Step 2: Use linear interpolation to complete the data. The calculation formula is as follows:

[0060]

[0061] Where y is the interpolated target value; y0 is the distance measurement value corresponding to the known point x0; y1 is the distance measurement value corresponding to the known point x1; x is the x-coordinate of the target interpolation point; x0 is the x-coordinate of the known point 1; and x1 is the x-coordinate of the known point 2.

[0062] Step 3: Use the quartile interval method to remove outlier data. The calculation formula is as follows:

[0063] IQR = Q3 - Q1,

[0064] Wherein, IQR is the interquartile range, which measures the dispersion of the data distribution; Q1 is the first quartile, representing the 25th percentile of the data; and Q3 is the third quartile, representing the 75th percentile of the data. The outlier judgment interval is Q1 - 1.5 × IQR and Q3 + 1.5 × IQR. Data points below the lower limit or above the upper limit of this interval are considered outliers.

[0065] Step 4: Apply pseudo-labels to the original data using the K-means clustering algorithm for training the SVM classification model. The goal of K-means is to minimize the sum of squared distances from each data point to its cluster center, iterating sequentially. The formula is as follows:

[0066]

[0067] Where J is the objective function for clustering; k is the number of clusters; C i Let x be the set of the i-th cluster; x is the set of the clusters belonging to C. i A specific data point in a cluster; μ i Let |x-μ be the centroid of the i-th cluster; i || 2 For data point x, the distance to cluster center μ i The square of the Euclidean distance. In each iteration, the cluster centers are updated according to the following formula:

[0068]

[0069] The K-means clustering algorithm iteratively optimizes the objective function. First, it performs initialization by randomly selecting k data points as initial centroids μ. i (i = 1, 2, ..., k); then, assign clusters. For each data point x, calculate its distance to the k centroids and assign it to the nearest cluster C. i ,in Next, the centroids are updated, and the centroid μ of each cluster is recalculated based on the assigned clusters. iFinally, repeat the cluster assignment and centroid update process until the centroid stops changing or the iteration limit is reached.

[0070] like Figure 1 As shown, the mean and standard deviation of UWB measurement data within a unit sliding window are calculated to construct a feature vector, which is then standardized using a feature scaler. The optimal combination of SVM model penalty parameters, kernel coefficients, and kernel function hyperparameters is optimized through grid search and cross-validation to complete SVM model training. The method specifically includes the following steps:

[0071] Step 5: To ensure sufficient model training, a simulation environment was constructed based on the real mitigation mapping, from which UWB simulation measurement data was generated and provided to the model for training. In the indoor environment, LOS and NLOS positioning errors follow Gaussian normal and exponential distributions, respectively. Based on whether the UWB signal propagation path between the tag and the base station is obstructed by obstacles in the simulation environment, the signal is divided into LOS and NLOS signals. According to measured data, the LOS positioning error accuracy is 5cm. Therefore, in the simulation, the LOS measurement data is introduced with an error having a mean of 0, a standard deviation of 0.05m, and following a normal distribution. The NLOS positioning error follows an exponential distribution, and its probability density function P(τ) is calculated as follows:

[0072]

[0073] Where τ NLOS =T1r ε ξ and T1 are the median of the delay spread at r = 10m, ε is a random scaling factor that adjusts the basic RMS delay, with a value ranging from 0.5 to 1, ξ is a log-normally distributed random variable with a standard deviation between 4 and 6 dB, and τ represents the root mean square delay spread.

[0074] Step 6: Divide the UWB measurement simulation dataset into a 70% training set and a 30% test set, and calculate the mean and standard deviation of the training set within the sliding window K. The mean and standard deviation are as follows:

[0075]

[0076] Among them, Mean K x is the average value within the window; i The i-th data point within the window; N represents the total number of data points within the window; SD K This represents the standard deviation within the window.

[0077] Step 7: The feature vectors calculated in Step 6 are standardized using the feature extractor StandardScaler. Training employs three-fold cross-validation with grid search. The training set is divided into three subsets; one subset is selected sequentially as the validation set, and the other two subsets are used as the training set for three independent training rounds to determine the optimal hyperparameter combination. The hyperparameter search range is as follows:

[0078]

[0079] Where C is the penalty parameter; γ is the kernel function coefficient; and Kernel is the kernel function. The appropriate setting of these parameters is crucial to the complexity and generalization ability of the SVM classification model.

[0080] like Figure 1 As shown, the training set is divided into NLOS and LOS training sets by the SVM classification model. The distance difference and time interval between different datasets within a unit sliding window after classification are calculated as additional features. Three features from the two classes trained in the SVM model are randomly selected as the basis features for the optimal split point. The optimal hyperparameter combination of the number of trees, maximum tree depth, minimum number of classification samples, and minimum number of node samples is optimized through grid search and cross-validation to complete the training of RF models for different datasets. The method specifically includes the following steps:

[0081] Step 8: Calculate the distance difference and timestamp time interval between the NLOS and LOS training sets within the sliding window K as additional features. The distance difference and time interval are as follows:

[0082] Δx k =x k -x k-1 ,

[0083] ΔT k =T k -T k-1 ,

[0084] Where, Δx k The distance difference within the window; x k ΔT represents the distance difference value for the k-th window. k T represents the time interval between timestamps within the window. k This represents the time interval value for the k-th window.

[0085] Step 9: Randomly select three features as the basis features for the best segmentation point. Training uses three-fold cross-validation for grid search to determine the optimal hyperparameter combination. The hyperparameter search range is as follows:

[0086]

[0087] Where T is the number of trees; Dmax N represents the maximum depth of the tree. split N is the minimum number of samples required for splitting; leaf This represents the minimum number of samples required for each leaf node. Properly setting these parameters is crucial to the complexity and generalization ability of the RF regression model.

[0088] Step 10: Evaluate the model using the test set and save the trained RF regression model.

[0089] like Figure 1 As shown, the calculation method of the error mitigation factor is determined based on the proportion of LOS data within a unit sliding window. When the proportion of LOS data is greater than zero, the ratio of the mean difference and standard deviation of LOS and NLOS measurement data is combined, and an exponential decay method is used to make the error mitigation factor more sensitive to changes in the degree of difference. A weighting parameter λ is introduced to control the combined influence of both on the final result. If the proportion of LOS is zero, the relative coefficient of variation of the NLOS data error is directly used, and an adjustment parameter γ is introduced to control the influence strength of the relative coefficient of variation. Finally, the error mitigation factor is output through an exponential function. The method specifically includes the following steps:

[0090] Step 11: When the proportion of LOS measurement data is greater than 0, first calculate the mean difference Δ. μ The difference between LOS and NLOS is measured and calculated as follows:

[0091]

[0092] Where, μ LOS The mean error of the LOS measurement data; μ NLOS This represents the mean error of the NLOS measurement data.

[0093] Step 12: Calculate the standard deviation ratio Δ σ The difference in scalability between LOS and NLOS measurement data is measured using the following formula:

[0094]

[0095] Where, σ LOS σ represents the standard deviation of the LOS measurement data. NLOS This represents the standard deviation of NLOS measurement data.

[0096] Step 13: Calculate the error mitigation factor using the ratio of the mean difference to the standard deviation, as shown in the following formula:

[0097]

[0098] Where L is the error mitigation factor; λ is the control factor used to adjust the influence intensity of the error mitigation factor.

[0099] Step 14: Combining the ratio of the mean difference to the standard deviation, we can derive the following formula:

[0100]

[0101] After extensive simulation experiments and optimization, λ was set to 1.5, and the error mitigation factor was calculated to be 0.76.

[0102] Step 15: When the proportion of LOS measurement data is 0, use the relative coefficient of variation of NLOS measurement data and calculate the error mitigation factor in the absence of LOS measurement data according to the formula:

[0103]

[0104] Here, γ is an adjustment parameter that controls the intensity of the error mitigation factor's influence. After extensive experimental optimization, the value of γ was set at 2.8, from which the error mitigation factor was calculated to be 0.61.

[0105] Step 16: UWB measurement data adjusted with error mitigation factor The calculation method is as follows:

[0106]

[0107] Where, x LOS To adjust the average value of LOS measurement data within the previous sliding window K; x NLOS To adjust the average value of NLOS measurements within the previous sliding window K; η LOS η represents the proportion of LOS measurement data within the sliding window k. LOS The calculation method is as follows:

[0108]

[0109] Where, N LOS N represents the number of LOS measurement data points within the sliding window K; Total This represents the total amount of UWB measurement data within the sliding window K.

[0110] like Figure 3 As shown, the localization method invented in this paper determines the appropriate localization method based on the proportion of LOS data after model classification and the total amount of data: if the proportion of LOS data exceeds 90% and the total amount of data is sufficient, localization is directly achieved using the Chan algorithm; if the LOS proportion is sufficient but the amount of data is insufficient, the model prediction is used to supplement the data before using the Chan algorithm for localization; if neither condition is met, corresponding error mitigation measures are implemented based on the presence of LOS data. Then, the measurement and prediction data are fitted using the least squares method, and the Chan algorithm is used for localization.

[0111] Step 17: The Chan algorithm calculation steps are as follows: calculate the distance difference r between base station i and base station j. ij :

[0112] r ij =d TW-TOFi -d TW-TOFj ,

[0113] Based on this, we can deduce that the target position satisfies the following formula:

[0114] ||pp i ||-||pp j ||=r ij ,

[0115] Where p is the coordinate to be determined; p i Given the coordinates of base station i, expanding the above formula, squaring it, and rearranging the squares, we get:

[0116]

[0117] Introduce auxiliary variable b = ||p|| 2 Linearizing the nonlinear equations yields the following matrix equations:

[0118] A·x=b,

[0119] Where x = [x, y, z, b] T , where is the vector of unknowns; A is the coefficient matrix; b is the vector of constant terms, and A has the following specific form:

[0120]

[0121] Solving the above linear equations using the least squares method yields a preliminary estimate of the target's position:

[0122] x est =(A T A) -1 A T b,

[0123] Where, x est We estimate the coordinates for the initial location. Using the initial value as the starting point for iteration, we further improve the accuracy through nonlinear optimization methods.

[0124] The innovative method for judging and mitigating ultra-wideband non-line-of-sight positioning errors of indoor mobile building robots, as described above, has many merits for technicians in the industry and is indeed a technological advancement.

[0125] The above describes the basic concept of this invention. However, within the technical field of this invention, with only basic knowledge, other operable embodiments of this invention can be improved. This invention seeks patent protection for the substantive technical solutions, and the scope of protection should include all variations possessing the aforementioned technical features.

[0126] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for judging and mitigating ultra-wideband non-line-of-sight positioning errors of indoor construction robots, characterized in that: This includes non-line-of-sight recognition methods and mitigation methods; The non-line-of-sight recognition method utilizes a trained support vector machine (SVM) classification model to classify preprocessed ultra-wideband measurement data within a unit sliding window to distinguish between line-of-sight and non-line-of-sight data. The training method for the SVM classification model involves calculating the mean and standard deviation of the ultra-wideband simulated measurement data within the unit sliding window to construct feature vectors, which are then standardized using a feature scaler. The optimal combination of the SVM model's penalty parameters, kernel coefficients, and kernel function is optimized through grid search and cross-validation to complete the SVM model training. The non-line-of-sight mitigation method determines the error mitigation factor calculation method based on the proportion of line-of-sight data within a unit sliding window. When the proportion of line-of-sight data is greater than zero, the ratio of the mean difference and standard deviation of line-of-sight and non-line-of-sight measurement data is combined, and an exponential decay form is used to make the error mitigation factor more sensitive to changes in the degree of difference. At the same time, a weighting parameter is introduced to control the combined influence of the two on the final result. If the proportion of line-of-sight is zero, the relative coefficient of variation of the non-line-of-sight data error is directly used, and an adjustment parameter is introduced to control the influence intensity of the relative coefficient of variation. Finally, the error mitigation factor is output through an exponential function. Due to non-line-of-sight signal propagation, some base stations may not be able to provide enough effective ranging data, resulting in the system being unable to obtain accurate positioning information from multiple base stations simultaneously, thus leading to insufficient data for positioning; a random forest model is used to predict ultra-wideband measurement data to supplement the completeness of the dataset. The non-line-of-sight mitigation method specifically includes the following steps: Ultrawideband measurement data adjusted by error mitigation factor The calculation method is as follows: ； in, To adjust the average value of the sight distance measurement data within the previous sliding window; To adjust the average value of non-line-of-sight measurement data within the previous sliding window; The proportion of distance measurement data within the sliding window; The calculation method is as follows: ； in, The number of line-of-sight measurement data within the sliding window; This represents the total amount of ultra-wideband measurement data within the sliding window. The error mitigation factor is calculated as follows: When the proportion of sight distance measurement data is greater than 0, the mean difference is first calculated to measure the difference between sight distance and non-sight distance. The calculation method is as follows: ； in, Difference between means; This represents the mean error of the stadia measurement data. This represents the mean error of non-line-of-sight measurement data; The standard deviation ratio is calculated to measure the difference in scalability between sight-based and non-sight-based measurements. The calculation method is as follows: ； in, This is the ratio of standard deviations; The standard deviation of the stadia measurement data; Standard deviation of non-line-of-sight measurement data; The error mitigation factor is calculated using the ratio of the mean difference to the standard deviation, as shown in the following formula: ； Where L is the error mitigation factor; This is a control factor used to adjust the strength of the error mitigation factor; combining the ratio of the mean difference to the standard deviation, we can derive the following formula: ； The value is set to 1.5, thus the error mitigation factor is calculated to be 0.

76. When the proportion of stadia measurement data is 0, the relative coefficient of variation of non-stadia measurement data is used, and the error mitigation factor in the absence of stadia measurement data is calculated according to the formula: ； in, To adjust parameters and control the intensity of the error mitigation factor; The value is set to 2.8, from which the error mitigation factor is calculated to be 0.61; Also includes: The appropriate localization method is determined based on the proportion of LOS data after model classification and the total amount of data: if the proportion of LOS data exceeds 90% and the total amount of data is sufficient, localization is directly achieved using the Chan algorithm; if the proportion of LOS data is sufficient but the amount of data is insufficient, the model is used to predict and supplement the data before using the Chan algorithm for localization; if neither of these conditions is met, appropriate error mitigation measures are implemented based on the presence of LOS data; then, the least squares method is used to fit the measured and predicted data, and the Chan algorithm is used for localization.

2. The method for judging and mitigating ultra-wideband non-line-of-sight positioning errors of indoor construction robots according to claim 1, characterized in that: The support vector machine model training process specifically includes the following steps: The ultra-wideband measurement simulation dataset is divided proportionally into... training set and For the test set, calculate the mean and standard deviation of the training set within the sliding window; the mean and standard deviation are as follows: ； Among them, Mean This represents the average value within the sliding window. Let be the i-th data point within the sliding window; N represents the total number of data points within the sliding window. The standard deviation within the sliding window; The calculated feature vectors are standardized by a feature extractor; training employs a grid search with three-fold cross-validation; the training set is divided into three subsets, with one subset selected sequentially as the validation set and the other two subsets used as the training set for three independent training rounds to determine the optimal hyperparameter combination; the hyperparameter search range is as follows: ； Where C is the penalty parameter; The coefficients are the kernel function coefficients; Kernel is the kernel function. A confusion matrix and learning curve are constructed using a test set to evaluate the model, and the trained support vector machine classification model is saved.

3. The method for judging and mitigating ultra-wideband non-line-of-sight positioning errors of indoor construction robots according to claim 1, characterized in that: The training process of the random forest model specifically includes the following steps: The distance difference and timestamp time interval between the non-line-of-sight and line-of-sight training sets within the sliding window are calculated as additional features, as shown in the following equation: ； in, The distance difference within the sliding window; This represents the distance difference value within the Kth sliding window; The time interval for timestamps within the sliding window; This represents the time interval value within the Kth sliding window; Three features are randomly selected as the basis for the optimal segmentation point. Training uses tri-fold cross-validation and grid search to determine the optimal hyperparameter combination. The hyperparameter search range is as follows: ； Where T is the number of trees; The maximum depth of the tree; The minimum number of samples required for splitting; Minimum number of samples for a leaf node; The model is evaluated using a test set, and the trained random forest regression model is saved.

Images