UWB NLOS identification and ranging error compensation method based on feature similarity
Through a dual input attention neural network based on feature similarity, the ranging error of UWB in the NLOS environment is identified and compensated, which solves the problem of reduced UWB ranging accuracy, and realizes efficient NLOS path recognition and error compensation, improving positioning Accuracy and model efficiency.
Patent Information
- Application Number
- CN202411102934.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-13
AI Technical Summary
UWB ranging technology is affected by factors such as multipath effect and signal attenuation in the NLOS environment, resulting in a decrease in the ranging accuracy. It is difficult for the existing technology to effectively identify the NLOS path and perform error compensation.
Using a dual input attention neural network based on feature similarity, the NLOS path is identified and error compensation is performed through the evaluation of channel characteristics and CIR similarity characteristics. The dual-layer attention mechanism is used to improve the processing efficiency of the model for input data, and the recognition is carried out. Switch between compensation tasks.
It significantly improves the ranging accuracy of UWB in the NLOS environment, reduces position estimation errors, and improves the utilization efficiency of the model, achieving accurate identification of NLOS paths and effective compensation of ranging errors.
Smart Images

Figure CN119004133A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of UWB ranging and positioning, and in particular relates to a UWB NLOS recognition and ranging error compensation method based on feature similarity. Background Art
[0002] Although UWB ranging technology is known for its high precision, it is still affected by various factors in actual applications, such as multipath effects, NLOS propagation, signal attenuation, clock synchronization errors, noise, hardware errors, temperature changes, bandwidth limitations, antenna position errors, and channel delay spread, resulting in ranging errors. By understanding and analyzing these error sources, corresponding compensation and correction measures can be taken to improve the accuracy and reliability of UWB ranging.
[0003] NLOS propagation refers to a situation where a wireless signal encounters an obstacle during transmission between the transmitter and the receiver, preventing the signal from reaching the receiver directly. This is in stark contrast to LOS propagation, where the signal is transmitted directly from the transmitter to the receiver without any physical obstacles. NLOS propagation is particularly common in urban environments, indoors, or other complex environments, and has a significant impact on the performance and reliability of wireless networks. Since signals need to bypass obstacles or propagate through reflection, diffraction, etc. under NLOS conditions, they usually experience significant attenuation, and the multipath effect is more prominent under such conditions, which may cause signal interference and affect the quality of the received signal.
[0004] When performing wireless positioning under NLOS conditions, the uncertainty of the signal propagation path may reduce the positioning accuracy. In order to improve the communication quality and reliability under NLOS conditions, advanced signal processing technologies such as multipath diversity reception, spatial diversity and frequency diversity can be used. In the face of the challenges brought by NLOS propagation, the design and optimization of wireless networks need to comprehensively consider a variety of technologies and strategies to overcome the problems in the NLOS environment, thereby improving the overall reliability and efficiency of wireless communications. Therefore, it is of great significance to identify the UWB NLOS path and compensate for the error. Summary of the invention
[0005] To solve the above problems, the present invention discloses a UWB NLOS recognition and ranging error compensation method based on feature similarity. Through a dual-input attention neural network, the input channel features and CIR similarity features are evaluated, and NLOS is first recognized and then compensated, which greatly improves the ranging accuracy of UWB in the NLOS environment and can significantly suppress the position estimation error of the target to be located. In addition, the model can also switch between the tasks of NLOS recognition and ranging error compensation, greatly improving the utilization efficiency of the model.
[0006] To achieve the above object, the technical solution of the present invention is as follows:
[0007] The UWB NLOS recognition and ranging error compensation method based on feature similarity includes the following steps:
[0008] A: Acquisition of channel characteristics. Ultra-wideband belongs to the field of radio frequency, and various channel characteristics can be extracted from ultra-wideband sample data. The extracted UWB channel characteristics not only include statistical information of observed distance values such as distance, maximum noise value, and noise standard deviation, but also include relevant characteristics of the first path. Among them, the first path is the shortest propagation path from the transmitter to the receiver, which usually represents the path with the shortest time required for signal propagation. In a complex environment, the first path is usually an NLOS path. In addition, the maximum noise value and the noise standard deviation are also clearly distinguished in the LOS / NLOS environment. These channel characteristics cooperate with the CIR similarity feature vector as input parameters to realize NLOS identification and ranging error compensation. Among them, CIR stands for channel impulse response, which is a measure of the channel's response to an impulse signal (i.e., an instantaneous unit energy pulse). It is an important parameter that characterizes the channel transmission characteristics and can reflect the changes in the signal during the channel propagation process, including NLOS, attenuation, delay, and multipath effects.
[0009] B: Calculation of CIR similarity feature vector. The calculation process of CIR similarity feature vector includes several key steps, as follows:
[0010] B1: Reference sample matching. Find the reference sample closest to the sample through distance calculation.
[0011] B2: Data filtering. The data is cleaned by averaging and threshold filtering to remove outliers and noise to improve data quality. The filtered data is used to establish the NLOS recognition and compensation model.
[0012] B3: Extract the CIR sequence near FP_Index from the measured channel characteristic data
[0013] B4: Perform digital filling. The CIR data near FP_Index is too sparse to describe the CIR shape similarity. It is necessary to fill some data equally between each two adjacent CIR data, and use the slope between the filled data to describe the shape similarity. Among them, FP_Index is the index of the first path peak, which refers to the position of the peak corresponding to the first signal path that reaches the receiver in the received signal in the signal data array.
[0014] B5: Use the CIR sequence before padding to calculate a similarity vector representing the CIR sequence value.
[0015] B6: Use the padded CIR sequence to calculate the similarity vector that indirectly describes the shape, concatenate the reference sampling error and the sample ranging error, and finally form a complete CIR similarity vector.
[0016] B7: Divide the processed dataset into training and test sets, and ensure that the data is evenly mixed when dividing the dataset so that it can be used in the model training and evaluation process.
[0017] Regarding the similarity features in the experimental data feature engineering, they mainly include cosine similarity, Spearman correlation coefficient, Pearson correlation coefficient, Canberra distance, standardized Euclidean distance, Jensen-Shannong divergence and Hellinger distance, etc. For the types of similarity features, combined with the actual scene selection, directly obtain the similarity vector between the CIR sequence near the FP_Index and the reference sample. As the numerical similarity information of the CIR vector. The shape difference characteristics of the prior reference CIR vector and the working sample CIR vector are also useful references for characterizing the LOS channel and the NLOS channel. For a dimensionless one-dimensional vector such as the CIR signal, its shape characteristics can be reflected using a slope vector. The specific calculation method is as follows:
[0018] (1) Cosine similarity, the cosine distance in n-dimensional space is shown as follows.
[0019]
[0020] Among them, x and y are two vectors to be compared. i ,y i is the i-th numerical component of the vector.
[0021] (2) Spearman correlation coefficient: The corresponding elements of the two observed variables are subtracted to obtain a difference d. The Spearman correlation coefficient can be calculated using the following formula.
[0022]
[0023] Among them, d i is the difference between the corresponding elements of the two vectors to be compared, and n is the dimension of the vector.
[0024] (3) Pearson correlation coefficient. The calculation formula of the Pearson correlation coefficient in n-dimensional space is as follows.
[0025]
[0026] Among them, σ(x) and σ(y) are the variances of the two vectors to be compared, and cov(x, y) is the covariance between the two vectors.
[0027] (4) Canberra distance. The Canberra distance in n-dimensional space is shown as follows.
[0028]
[0029] Among them, x i ,y i are the i-th numerical components of the vectors to be compared.
[0030] (5) Standard Euclidean similarity. The calculation formula of standard Euclidean similarity can be expressed as follows.
[0031]
[0032] Where d(x, y) is the standard Euclidean distance between two vectors.
[0033] (6) Jensen-Shanon divergence. The Jensen-Shanon divergence can effectively measure the similarity between two probability distributions, overcome the asymmetry problem existing in KL divergence, and provide a standard way to quantify the difference between probability distributions. The following formula is the calculation formula of Jensen-Shanon divergence:
[0034]
[0035] Among them, KL is the KL divergence between the two directions.
[0036] (7) Hellinger distance is a method used to measure the difference between two probability distributions. It is used to measure the similarity between two probability distributions. For discrete probability distributions, Hellinger distance is defined as shown below.
[0037]
[0038] Where p and q are two probability distributions defined on the same event space, and p = {p 1 ,p 2 ,...,p n}, q={q 1 ,q 2 ,...,q n}.
[0039] C: Construction of NLOS identification and error compensation model
[0040] Based on the design of CIR similarity features, a two-layer attention neural network is built. The two input layers input the channel feature vector and the CIR similarity feature vector respectively, and a NLOS recognition deep learning model is constructed. In order to avoid the problem that the same attention mechanism has poor evaluation effect on different types of inputs, the two groups of inputs are evaluated by the attention mechanism separately. The purpose is to improve the model's recognition and processing efficiency of key information in the input data.
[0041] Adjust the parameter structure and build an error compensation model. The adjusted parameters include the optimizer and loss function to achieve the switch between the classification model and the regression model. The difference between the two lies in the output layer structure. The classification model output layer has two neurons and two bias items; the compensation model output layer has only one neuron and one bias item.
[0042] Step D: The NLOS recognition deep learning model and error compensation model are used to realize the recognition of UWB NLOS path and the ranging error compensation.
[0043] The beneficial effects of the present invention are:
[0044] (1) The present invention uses a dual-input attention neural network to perform feature evaluation on the input channel features and CIR similarity features, first identify and then compensate for NLOS, which greatly improves the ranging accuracy of UWB in the NLOS environment and can significantly suppress the position estimation error of the target to be located.
[0045] (2) The present invention also realizes the recognition and error compensation functions of UWB NLOS by adjusting the model parameters, thereby improving the utilization efficiency of the model and reducing the size of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 This is a schematic diagram of the overall framework of the UWB NLOS identification and error compensation method proposed in the present invention.
[0047] Figure 2 The present invention proposes a schematic diagram of the overall process of the UWB NLOS recognition model.
[0048] Figure 3 The present invention proposes a schematic diagram of the overall flow of the UWB NLOS error compensation model.
[0049] Figure 4 This is a schematic diagram of the UWB NLOS identification and error compensation implementation method proposed in this patent. DETAILED DESCRIPTION
[0050] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.
[0051] like Figure 1 As shown, the NLOS identification and ranging error compensation algorithm based on UWB channel characteristics and CIR similarity characteristics described in the present invention includes the following key steps:
[0052] (1) Use UWB sampling equipment to obtain distance observation values, RF channel characteristics and CIR signals (including reference CIR information and sampled CIR information).
[0053] (2) Calculate the channel similarity between the reference CIR and the sampled CIR and construct a channel similarity feature vector
[0054] (3) Based on the pattern input of RF channel characteristics and channel similarity feature vectors, a deep learning model for NLOS recognition is constructed.
[0055] (4) Adjust the parameter structure and transform the NLOS identification model into an error compensation model.
[0056] Generally speaking, the invention mainly includes two modules: obtaining input parameters and building a model.
[0057] The specific implementation steps include:
[0058] Step A: Experimental data feature engineering, extracting channel features and calculating similarity feature vectors for the collected experimental data.
[0059] The specific steps of experimental data feature engineering are as follows:
[0060] Step A1: Reference sample matching: by distance calculation, find the reference sample closest to the sample, so that the channel characteristics and CIR sequence of the sample are paired and compared with the appropriate reference sample.
[0061] Step A2: Data filtering, cleaning the data through average filtering and threshold filtering, removing outliers and noise to improve data quality
[0062] Step A3: Acquisition of channel characteristics. These channel characteristics can be directly acquired from the data collected by the UWB module.
[0063] Step A4: Acquisition of CIR sequence: extracting the CIR sequence near FP_Index from the measured channel characteristic data.
[0064] Step A5: Digital filling: The CIR data near FP_Index is too sparse to describe the CIR shape similarity. It is necessary to fill some data at equal intervals between each two adjacent CIR data, and use the slope between the filled data to describe the shape similarity.
[0065] Step A6: Calculate the similarity feature vector. Use the CIR sequence before padding according to the similarity feature calculation method mentioned above to calculate the similarity vector that represents the numerical value of the CIR sequence. Then use the CIR sequence after padding to calculate the similarity vector that indirectly describes the shape. The error of the reference sampling and the ranging error of the sample are concatenated to finally form the CIR similarity vector.
[0066] Step A7: Dataset division: divide the processed dataset into training set and test set. When dividing the dataset, ensure that the data is evenly mixed so that it can be used in the model training and evaluation process. In this experiment, the dataset division ratio is 1:1.
[0067] Step B: Construction of NLOS identification and error compensation model. The two-layer attention fusion network consists of 7 modules, including input layer, linear combination and batch normalization, attention mechanism, feature merging, deep feature extraction, dropout layer and output layer. It contains two input layers, which input channel feature vector and similarity feature vector respectively. In order to avoid the problem that the same attention mechanism has poor evaluation effect on different types of inputs, the two groups of inputs are evaluated by the attention mechanism separately.
[0068] Step C: After completing the construction of the NLOS identification and error compensation model, deploy the model to the edge to realize the identification of the UWB NLOS path and the ranging error compensation.
[0069] In step A, regarding the acquisition of channel characteristics, ultra-wideband belongs to the radio frequency field. The channel characteristics can extract various channel characteristics from ultra-wideband sample data. The extracted UWB channel characteristics not only include statistical information of observed distance values such as distance, maximum noise value, and noise standard deviation, but also include relevant characteristics of the first path. Among them, the first path is the shortest propagation path from the transmitter to the receiver, which usually represents the path with the shortest time required for signal propagation. In a complex environment, the first path is usually an NLOS path. In fact, the available first path characteristics, namely the first path peak and the first path strength, are very important for evaluating the ultra-wideband ranging error. In addition, channel characteristics such as maximum noise value and noise standard deviation are also clearly distinguished in LOS / NLOS environments.
[0070] The similarity features in feature engineering mainly include cosine similarity, Spearman correlation coefficient, Pearson correlation coefficient, Canberra distance, standardized Euclidean distance, Jensen-Shannong divergence and Hellinger distance. The similarity vector between the CIR sequence near FP_Index and the reference sampling can be directly obtained. The shape difference characteristics of the prior reference CIR vector and the working sampling CIR vector are also useful references for characterizing LOS channels and NLOS channels. For a dimensionless one-dimensional vector such as a CIR signal, its shape characteristics can be reflected using a slope vector. The specific calculation method is as described above. These similarity features reflect the severity of NLOS interference to a certain extent.
[0071] In step B, the core of the model building lies in the attention mechanism, whose purpose is to improve the model's recognition and processing efficiency of key information in the input data. Figure 2 and Figure 3The flowcharts for NLOS and ranging error compensation are shown respectively. Recognition For each batch of data that has been processed by linear transformation and batch normalization, a fully connected layer is used to calculate the attention weights separately. In the classification task, the SoftMax activation function is used to ensure that the sum of the feature weights is 1, so that it can be interpreted as a probability distribution. For the LOS / NLOS ranging error compensation task, the linear rectifier unit activation function is used to generate non-negative attention weights. In this way, the calculated attention weights are multiplied by the corresponding linear transformation output to achieve the purpose of feature weighting, effectively adjusting the influence of each feature on the model output. This not only increases the model's ability to handle complex data, but also improves its interpretability. The attention weights directly show the contribution of each feature to the prediction result.
[0072] After completing feature evaluation, feature fusion is required. The goal of feature fusion is to integrate the information of two independent input sources and fully understand the input data by effectively fusing features. This not only improves the model's ability to represent data, but also lays a solid foundation for in-depth feature extraction and final decision-making. In this model, the implementation of feature fusion mainly goes through the following steps:
[0073] (1) Calculate attention weights for channel features and CIR similarity features respectively, and apply these weights to the corresponding features to highlight the features that are severely affected by NLOS. This operation is implemented through multiply, allowing the model to automatically focus on the most important features for NLOS recognition. The multiply function refers to the operation of element-by-element multiplication of the input tensor, which is often used to implement certain activation functions, weight sharing, or feature combination. Through element-by-element multiplication, different features or signals are effectively combined, thereby improving the expressiveness of the model.
[0074] (2) Use the concatenate operation to combine the weighted channel features and similarity features to create a comprehensive feature representation. The concatenate function is used to connect multiple tensors along a specified axis. This operation can merge different features or data into a larger tensor, thereby increasing the dimension of information. This function is often used to merge features from different layers, or to combine information from different sources when processing multimodal data. This operation helps the model better learn and express complex relationships and features.
[0075] In summary, the UWB NLOS recognition and ranging error compensation algorithm based on feature similarity proposed in the present invention makes full use of the difference between the CIR sequence peak in the NLOS environment and the CIR sequence peak in the LOS environment. Based on this, a set of CIR similarity vector calculation methods are designed. Compared with directly inputting the CIR sequence, the input parameter amount of the model is reduced. The model can also switch between the tasks of NLOS recognition and ranging error compensation, which greatly improves the utilization efficiency of the model. At the same time, the model first recognizes and then compensates for NLOS, which greatly improves the ranging accuracy of UWB in the NLOS environment, provides a basis for high-precision positioning of UWB, and can significantly suppress the position estimation error of the target to be located.
[0076] After the above steps, the NLOS identification and error compensation model can be deployed on the edge device to perform specific tasks. The process is as follows: Figure 4 shown.
[0077] It should be noted that the above content only illustrates the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications all fall within the protection scope of the claims of the present invention.
Claims
1. A UWB NLOS recognition and ranging error compensation method based on feature similarity, characterized in that: Through a dual-input attention neural network, the input channel characteristics and CIR similarity characteristics are evaluated, thereby realizing the identification of UWB NLOS paths and ranging error compensation; when designing the CIR similarity feature vector, based on the obvious differences in CIR sequences under LOS and NLOS environments, the similarity is calculated from the CIR numerical difference and shape difference respectively. The designed CIR similarity vector completely evaluates the NLOS interference degree from the aspects of numerical value and shape, which serves as an important basis for NLOS identification and error compensation.
2. The UWB NLOS identification and ranging error compensation method based on feature similarity according to claim 1, characterized in that: The position of FP_Index in CIR is directly related to the time delay of signal propagation. The NLOS interference level is evaluated based on the peak arrival time and peak height of the CIR sequence.
3. The UWB NLOS identification and ranging error compensation method based on feature similarity according to claim 1, characterized in that: When calculating the CIR similarity feature vector, select the reference sample close to the ideal LOS environment. When calculating the CIR similarity feature, select the reference sample with the closest distance for similarity vector comparison calculation.
4. The UWB NLOS identification and ranging error compensation method based on feature similarity according to claim 1, characterized in that: The CIR difference between LOS and NLOS environments does not always exist. The difference is reflected in some CIR sequences near FP_Index. When using the CIR sequence to calculate the similarity feature vector, the original CIR sequence is truncated and the CIR sequence with a large difference is retained.
5. The UWB NLOS identification and ranging error compensation method based on feature similarity according to claim 1, characterized in that: The CIR similarity feature vector includes similar features including a numerical similarity vector and a computable shape similarity vector between the sample CIR sequence and the reference sample CIR sequence; by directly calculating the partial similarity index between the CIR sequence near FP_Index and the reference sample CIR sequence, a similarity vector reflecting the numerical aspects of the sample and reference sample CIR sequences is obtained.
6. The UWB NLOS identification and ranging error compensation method based on feature similarity according to claim 5, characterized in that: The shape similarity vector can be calculated, the shape difference characteristics of the prior reference CIR vector and the working sampled CIR vector, for the dimensionless one-dimensional vector of the CIR signal, its shape characteristics are used Slope vector reflection: By visualizing the original CIR sequence, a vector reflecting the CIR shape similarity is constructed; when constructing the shape similarity vector, CIR sequence filling and graphic slope calculation are used to improve the degree of shape restoration.
7. The UWB NLOS identification and ranging error compensation method based on feature similarity according to claim 1, characterized in that: The two-layer attention NLOS recognition and error compensation neural network has two input layers that input channel features and CIR similarity feature vectors respectively, which enter the neural network after being evaluated by the attention mechanism, thereby realizing the recognition of NLOS paths; The specific implementation steps include: Step A: Experimental data feature engineering, extracting channel features and calculating CIR similarity feature vectors for the collected experimental data; The specific steps of experimental data feature engineering are as follows: Step A1: Reference sample matching, find the reference sample closest to the sample through distance calculation; Step A2: Data filtering, cleaning the data through average filtering and threshold filtering, removing outliers and noise to improve data quality; Step A3: Acquisition of channel characteristics, which can be directly obtained from the data collected by the UWB module; Step A4: Acquisition of CIR sequence: extracting CIR sequences near FP_Index from the measured channel characteristic data. The difference of NLOSCIR sequences is mainly reflected by these CIR sequences. Step A5: Digital filling: The CIR data near FP_Index is too sparse to describe the CIR shape similarity. It is necessary to fill some data equally between each two adjacent CIR data, and use the slope between the filled data to describe the shape similarity. Step A6: Calculate the similarity feature vector. Use the CIR sequence before padding to calculate the similarity vector that represents the CIR sequence value. Then use the CIR sequence after padding to calculate the similarity vector that indirectly describes the shape. The error of the reference sampling and the ranging error of the sample are concatenated to form the final input similarity vector. Step A7: Dataset division: divide the processed dataset into training set and test set. When dividing the dataset, ensure that the data is evenly mixed so that it can be used in the model training and evaluation process. In this training process, the dataset division ratio is 1:
1. Step B: Based on the design of CIR similarity features, a UWB NLOS recognition and ranging error compensation algorithm is constructed. The algorithm builds a two-layer attention neural network. The two input layers input the channel feature vector and the CIR similarity feature vector respectively. The two sets of inputs are evaluated by the attention mechanism respectively, and the features are extracted through the deep neural network to complete the NLOS recognition; the channel feature vector and CIR similarity feature vector of the identified NLOS type data are used as input parameters to enter the ranging error compensation neural network to complete the compensation task.
8. The UWB NLOS recognition and ranging error compensation algorithm based on feature similarity according to claim 7, characterized in that: The NLOS recognition algorithm and the error compensation algorithm use the same neural network structure, and by adjusting the parameters, they can flexibly switch between the classification and compensation tasks.
Citation Information
Patent Citations
Ultra-wideband non-line-of-sight signal identification method and device and computer equipment
CN115293246A
UWB NLOS propagation error suppression method based on deep learning
CN116321423A
Positioning method based on mixed fingerprints and line-of-sight / non-line-of-sight recognition
CN117278934A
Cited By
High-precision anti-interference distance measurement method based on UWB
CN119805429A
Non-line-of-sight classification and error compensation method and system based on double convolutional neural networks
CN121093166A
Non-line-of-sight classification and error compensation method and system based on double convolution neural network
CN121093166B
