Pedestrian GNSS / INS self-adaptive integrated navigation method based on scene classification
By using GRU-based scene classification and an improved adaptive Kalman filter algorithm, combined with a window smoothing method, the problems of poor positioning accuracy and power consumption optimization in pedestrian navigation under complex environments are solved, achieving high-precision navigation and positioning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-16
- Publication Date
- 2026-04-03
AI Technical Summary
In existing technologies, pedestrian navigation suffers from poor positioning accuracy and power consumption optimization in complex environments, and traditional extended Kalman filter algorithms cannot effectively suppress error accumulation.
A GRU-based scene classification model is used to classify raw GNSS data. Combined with IMU data, an improved adaptive Kalman filter algorithm and window smoothing method are used to adaptively adjust the GNSS measurement variance matrix through scene, reduce the weight of the abnormal observation model, and improve positioning accuracy.
It effectively improves the positioning accuracy of pedestrian navigation in complex environments, reduces positioning errors, lowers power consumption, and enhances the system's anti-interference capability.
Smart Images
Figure CN121783128A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of navigation and positioning technology, and in particular to a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method. Background Technology
[0002] In recent years, navigation and positioning services have been widely used in both military and civilian fields, gradually becoming an increasingly important part of people's lives. Accurate and continuous pedestrian location information is widely used in professional applications such as armed police duty, field patrols, and pipeline maintenance. Satellite navigation systems provide good location services for users in open outdoor environments; however, the system cannot function well when GNSS signal quality is unreliable. Other wireless positioning technologies, such as wireless LAN, ultra-wideband, and radio frequency identification (RFID), can directly provide location information. However, these technologies require the prior deployment of some infrastructure, and their signals are easily interfered with. Although cameras, lidar, and other technologies can effectively improve the positioning accuracy of pedestrian positioning systems, they are challenging to implement for low-overhead platforms and are susceptible to environmental factors.
[0003] In traditional multi-source fusion navigation algorithms, the traditional Extended Kalman Filter (EKF) algorithm is widely used due to its simplicity and versatility, but it does not have the ability to suppress error accumulation. Summary of the Invention
[0004] This invention provides a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method to address the shortcomings of existing pedestrian positioning systems in complex environments, such as poor accuracy and difficulty in power consumption optimization.
[0005] In a first aspect, the present invention provides a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method, comprising: Collect raw GNSS data and preprocess the raw GNSS data to obtain preprocessed GNSS data; The raw GNSS data is classified using a GRU-based scene classification model to obtain scene classification results; IMU data is acquired, and combined with the preprocessed GNSS data and the scene classification results, the integrated navigation result is obtained based on the improved adaptive Kalman filter algorithm.
[0006] According to the present invention, a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method is provided, which involves collecting raw GNSS data and preprocessing the raw GNSS data to obtain preprocessed GNSS data, including: The raw GNSS data includes pseudorange, carrier wave, number of satellites, PDOP, and signal-to-noise ratio; Significant noise signals are removed from the raw GNSS data to obtain the GNSS positioning results.
[0007] According to the present invention, a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method is provided, which uses a GRU-based scene classification model to classify the raw GNSS data to obtain scene classification results, including: The raw GNSS data is input into a GRU-based scene classification model to perform GRU scene classification and obtain the variance R matrix estimate. The scene classification results include open environment, tree-lined road environment, high-rise building environment, and bridge arch environment.
[0008] According to the present invention, a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method acquires IMU data, combines the preprocessed GNSS data and the scene classification results, and obtains integrated navigation results based on an improved adaptive Kalman filter algorithm, including: Error compensation is performed on IMU data by sensors, followed by mechanical orchestration, zero-velocity detection, and ZUPT. The GNSS measurement variance matrix in the Kalman filter is adaptively adjusted according to the scene. An improved window smoothing method is adopted, which uses static and dynamic weights to process the residual information vector of the time window, calculates the covariance matrix of the state prediction information, and obtains the final integrated navigation result.
[0009] According to the present invention, a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method is provided, which adaptively adjusts the GNSS measurement variance matrix in the Kalman filter according to the scene, including: When the detected scenario is a tall tree-lined road or a tunnel, using information from previous multiple epochs, the predicted residual information vector for GNSS / INS integrated navigation based on a Kalman filter framework is... and the corresponding theoretical covariance Represented as:
[0010]
[0011] in, and These represent the measurement matrix and the prediction variance matrix, respectively. and Represents the actual GNSS observation vector and the state estimate at time k. Represents the observation noise covariance matrix. This represents the variance matrix of the predicted residuals.
[0012] The present invention provides a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method, which employs an improved window smoothing method, utilizes static and dynamic weights to process the residual information vector of the time window, and calculates the covariance matrix of the state prediction information, including: Set the weight of the i-th sample at time k:
[0013]
[0014]
[0015] in, and Let these represent the static and dynamic weights of the i-th sample at time k, respectively. This is a smoothing parameter, where N represents the sliding window size; Obtain the predicted covariance matrix: .
[0016] According to the scene classification-based pedestrian GNSS / INS adaptive integrated navigation method provided by the present invention, when the actual noise is relatively small in the predicted covariance matrix, If the value is negative, it cannot be used as the covariance matrix of the observation vector. A statistic is expressed as: if the error of the dynamic model is large, the absolute value of the covariance of the prediction residual is also large due to the cumulative effect of the error statistic. Based on the scene classification results and statistical data, the following judgment conditions are determined:
[0017] in This represents the GNSS measurement variance that can be estimated based on changes in the scene. and These represent the minimum and maximum values of the preset GNSS measurement variance, respectively.
[0018] Secondly, the present invention also provides a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation system, comprising: The preprocessing module is used to collect raw GNSS data and preprocess the raw GNSS data to obtain preprocessed GNSS data. The classification module is used to classify the raw GNSS data using a GRU-based scene classification model to obtain scene classification results; The positioning module is used to acquire IMU data, combine the preprocessed GNSS data and the scene classification results, and obtain the integrated navigation result based on the improved adaptive Kalman filter algorithm.
[0019] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the scene classification-based pedestrian GNSS / INS adaptive integrated navigation method as described above.
[0020] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the scene classification-based pedestrian GNSS / INS adaptive integrated navigation method as described above.
[0021] This invention provides a scene-classification-based adaptive GNSS / INS integrated navigation method for pedestrians. It classifies scenes using a GRU-based scene classification model and then proposes an adaptive robust factor based on scene classification to detect observation model anomalies that may occur under complex conditions during integrated navigation. This invention can be applied to mainstream embedded platforms and effectively solves the positioning problem of multi-source fusion positioning systems in complex scenes. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0023] Figure 1 This is one of the flowcharts of the scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method provided by the present invention; Figure 2 This is the second flowchart of the scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method provided by the present invention; Figure 3 This is a comparison diagram of different scenario classification methods provided by the present invention; Figure 4 This is a comparison diagram of the positioning trajectories of different solutions provided by this invention; Figure 5 This is a schematic diagram of the structure of the scene-classification-based pedestrian GNSS / INS adaptive integrated navigation system provided by the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0025] To address the limitations of existing technologies, this invention proposes a scene-classification-based adaptive GNSS / INS fusion positioning algorithm for pedestrians. This algorithm can solve the problem of poor positioning accuracy in pedestrian integrated navigation positioning. It classifies the scene in which the pedestrian is located using a GRU-based scene classification algorithm and designs a scene-based adaptive robust algorithm to reduce the impact of abnormal GNSS observations on the positioning accuracy of integrated navigation.
[0026] Figure 1 This is one of the flowcharts of scene-classified pedestrian GNSS / INS adaptive integrated navigation provided in this embodiment of the invention, such as... Figure 1 As shown, it includes: Step 100: Collect raw GNSS data and preprocess the raw GNSS data to obtain preprocessed GNSS data; Step 200: Classify the raw GNSS data using a GRU-based scene classification model to obtain scene classification results; Step 300: Acquire IMU data, combine the preprocessed GNSS data and the scene classification results, and obtain the integrated navigation results based on the improved adaptive Kalman filter algorithm.
[0027] The purpose of this invention is to develop an adaptive GNSS / INS fusion positioning algorithm for pedestrians in complex scenarios, mitigating the impact of poor GNSS signal strength on positioning results. This positioning scheme addresses the problem of poor positioning accuracy in pedestrian GNSS / INS integrated navigation positioning. Specifically, it classifies scenes using a neural network-based scene classification method and employs an improved adaptive Kalman filter algorithm based on the scene classification results to enhance the system's positioning accuracy. This positioning scheme can be applied to various platforms, including vehicle-mounted, aircraft-mounted, and pedestrian-based systems.
[0028] Specifically, such as Figure 2As shown, firstly, the number of satellites, Position Dilution of Precision (PDOP) value, and signal-to-noise ratio are collected using a GNSS / BeiDou chip. The collected data is then used to classify the pedestrian's current scene using a GRU neural network. Finally, an improved scene-based adaptive Kalman filter algorithm is constructed to suppress the influence of abnormal GNSS information on the observation model, thereby improving positioning accuracy.
[0029] This scene-classification-based adaptive GNSS / INS integrated navigation algorithm for pedestrians employs an adaptive Kalman filter during localization, effectively reducing the weight of outlier observation models to minimize their impact on the system's localization results. When the detected scene is classified as a tall tree-lined road or a tunnel, information from previous multiple epochs is used to predict the residual information vector for GNSS / INS integrated navigation based on the Kalman filter framework. and the corresponding theoretical covariance It can be represented as:
[0030]
[0031] in, and These represent the measurement matrix and the prediction variance matrix, respectively. and Represents the actual GNSS observation vector and the state estimate at time k. Represents the observation noise covariance matrix. This represents the variance matrix of the predicted residuals.
[0032] Meanwhile, this invention proposes an improved window smoothing method, which uses static and dynamic weights to process the residual information vector of the time window and calculates the covariance matrix of the state prediction information. The weight of the i-th sample at time k can be expressed as follows:
[0033]
[0034]
[0035] in, and Let these represent the static and dynamic weights of the i-th sample at time k, respectively. Here, is the smoothing parameter, and N represents the sliding window size. Therefore, the prediction covariance matrix can be obtained as follows:
[0036] Obviously, when the actual noise is relatively small, A negative value cannot be used as the covariance matrix of the observation vector. To address this issue, a statistic is proposed: if the error of the dynamic model is large, the absolute value of the covariance of the prediction residuals will also be large due to the cumulative effect of the error statistic. Based on scene classification results and statistical data, the following judgment condition is proposed to address the above problems:
[0037] in This represents the GNSS measurement variance that can be estimated based on changes in the scene. and These represent the minimum and maximum values of the preset GNSS measurement variance, respectively.
[0038] The localization process of this scene classification-based pedestrian GNSS / INS adaptive integrated navigation algorithm mainly includes three steps: data acquisition and data preprocessing, scene classification, and fusion localization.
[0039] 1) Data Acquisition and Preprocessing The accuracy of scene classification is greatly affected by noise and interference in the satellite data collected by GNSS chips. Firstly, significant noise signals in the GNSS data are eliminated to improve the accuracy of positioning results.
[0040] 2) GRU-based scene classification To more accurately determine the current scene type, the algorithm uses a GRU-based scene classification model to divide the actual scene into four types: open environment, tree-lined road environment, high-rise building environment, and bridge hole environment.
[0041] 3) Adaptive fusion positioning of GNSS and INS By analyzing the signal characteristics of different scene types and considering the stability of the algorithm, the GNSS variance matrix under different scenes is determined by calculating the residual matrix. For open environments, GNSS positioning performance is good, and the variance matrix only needs to be set to the minimum. For tree-lined roads and high-rise building environments, adaptive filtering is required, and the GNSS variance needs to be set according to different attenuation factors. For bridge underpass environments, GNSS positioning performance is the worst, and the GNSS variance is set to the maximum value.
[0042] To test the classification accuracy of the GRU network in this scheme, a multi-source fusion navigation module was attached to a pedestrian's foot to collect GNSS data. The scene classification model of this scheme was then compared with classifiers based on recurrent neural networks (RNN), convolutional neural networks (CNN), and support vector machines (SVM). Figure 3The results of four classification methods are presented, and Table 1 shows the classification accuracy of different methods. Table 1 shows that the GRU network achieves an accuracy of over 93% in open, tunnel, and high-rise building environments, and an accuracy of 84% for tree-lined roads. Figure 3 As can be seen, the GRU network has the best classification performance, while the SVM network makes the most classification errors, even failing to properly identify open environments. All three classification networks suffer from confusing high-rise buildings with tree-lined avenues. Overall, the GRU classifier of this invention achieves better classification accuracy.
[0043] To verify the positioning performance of this invention, the measured positioning results were compared with four other schemes: ZUPT, GNSS, traditional GNSS / INS, and adaptive filtering. Figure 4 The positioning trajectories estimated using different methods are shown. From Figure 4 As can be seen, due to heading drift, the trajectory estimated by the ZUPT method gradually deviates from the true trajectory over time. GNSS signals are often affected by environmental interference, resulting in sometimes significant positioning errors. Traditional GNSS / INS fusion can effectively improve positioning results, but positioning accuracy drops drastically when GNSS signals are poor. This invention effectively addresses positioning issues caused by poor GNSS signals and comprehensively improves positioning results after GNSS and inertial guidance fusion.
[0044] Table 1 shows the maximum error and root mean square error (RMSE) of the trajectory estimated by different methods. As can be seen from Table 3, the ZUPT method has a maximum error of 39.42 meters and an RMSE of 16.92 meters. The GNSS method has a maximum deviation of 17.53 meters and an RMSE of 4.99 meters. The RMSEs of the traditional GNSS / INS method, the adaptive GNSS / INS method, and the proposed method are 3.65 meters, 3.22 meters, and 2.67 meters, respectively. Compared with the other four methods, the RMSE of this invention is reduced by 84.21%, 46.49%, 26.84%, and 17.08%, respectively. Therefore, the positioning method of this invention can effectively reduce the positioning error of pedestrians.
[0045] Table 1. Statistical analysis of classification results for different scenarios
[0046] Table 2. Statistics of positioning results from different positioning methods
[0047] The following describes the scene-classification-based pedestrian GNSS / INS adaptive integrated navigation system provided by the present invention. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation system described below can be referred to in correspondence with the scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method described above.
[0048] Figure 5 This is a schematic diagram of the structure of the scene-classification-based pedestrian GNSS / INS adaptive integrated navigation system provided in an embodiment of the present invention, as shown below. Figure 5 As shown, it includes: a preprocessing module 51, a classification module 52, and a positioning module 53, wherein: The preprocessing module 51 is used to collect raw GNSS data and preprocess the raw GNSS data to obtain preprocessed GNSS data; the classification module 52 is used to classify the raw GNSS data using a GRU-based scene classification model to obtain scene classification results; the positioning module 53 is used to acquire IMU data, and combine the preprocessed GNSS data and the scene classification results to obtain integrated navigation results based on an improved adaptive Kalman filter algorithm.
[0049] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method. This method includes: acquiring raw GNSS data; preprocessing the raw GNSS data to obtain preprocessed GNSS data; classifying the raw GNSS data using a GRU-based scene classification model to obtain scene classification results; acquiring IMU data; and combining the preprocessed GNSS data and the scene classification results with an improved adaptive Kalman filter algorithm to obtain integrated navigation results.
[0050] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0051] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the scene classification-based pedestrian GNSS / INS adaptive integrated navigation method provided by the above methods. The method includes: acquiring raw GNSS data; preprocessing the raw GNSS data to obtain preprocessed GNSS data; classifying the raw GNSS data using a GRU-based scene classification model to obtain scene classification results; acquiring IMU data; and combining the preprocessed GNSS data and the scene classification results to obtain integrated navigation results based on an improved adaptive Kalman filter algorithm.
[0052] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0053] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method, characterized in that, include: Collect raw GNSS data and preprocess the raw GNSS data to obtain preprocessed GNSS data; The raw GNSS data is classified using a GRU-based scene classification model to obtain scene classification results; IMU data is acquired, and combined with the preprocessed GNSS data and the scene classification results, the integrated navigation result is obtained based on the improved adaptive Kalman filter algorithm.
2. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method according to claim 1, characterized in that, Acquire raw GNSS data, and preprocess the raw GNSS data to obtain preprocessed GNSS data, including: The raw GNSS data includes pseudorange, carrier wave, number of satellites, PDOP, and signal-to-noise ratio; Significant noise signals are removed from the raw GNSS data to obtain the GNSS positioning results.
3. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method according to claim 1, characterized in that, The raw GNSS data is classified using a GRU-based scene classification model to obtain scene classification results, including: The raw GNSS data is input into a GRU-based scene classification model to perform GRU scene classification and obtain the variance R matrix estimate. The scene classification results include open environment, tree-lined road environment, high-rise building environment, and bridge arch environment.
4. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method according to claim 1, characterized in that, IMU data is acquired, and combined with the preprocessed GNSS data and the scene classification results, based on the improved adaptive Kalman filter algorithm, the integrated navigation results are obtained, including: Error compensation is performed on IMU data by sensors, followed by mechanical orchestration, zero-velocity detection, and ZUPT. The GNSS measurement variance matrix in the Kalman filter is adaptively adjusted according to the scene. An improved window smoothing method is adopted, which uses static and dynamic weights to process the residual information vector of the time window, calculates the covariance matrix of the state prediction information, and obtains the final integrated navigation result.
5. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method according to claim 4, characterized in that, The GNSS measurement variance matrix in the Kalman filter is adaptively adjusted according to the scene, including: When the detected scenario is a tall tree-lined road or a tunnel, using information from previous multiple epochs, the predicted residual information vector for GNSS / INS integrated navigation based on a Kalman filter framework is... and the corresponding theoretical covariance Represented as: in, and These represent the measurement matrix and the prediction variance matrix, respectively. and Represents the actual GNSS observation vector and the state estimate at time k. Represents the observation noise covariance matrix. This represents the variance matrix of the predicted residuals.
6. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method according to claim 5, characterized in that, An improved window smoothing method is employed, utilizing static and dynamic weights to process the residual information vector of the time window, and calculating the covariance matrix of the state prediction information, including: Set the weight of the i-th sample at time k: in, and Let these represent the static and dynamic weights of the i-th sample at time k, respectively. This is a smoothing parameter, where N represents the sliding window size; Obtain the predicted covariance matrix: 。 7. The scene-classification-based pedestrian GNSS / INS adaptive integrated navigation method according to claim 6, characterized in that, In the predicted covariance matrix, when the actual noise is relatively small, If the value is negative, it cannot be used as the covariance matrix of the observation vector. A statistic is expressed as: if the error of the dynamic model is large, the absolute value of the covariance of the prediction residual is also large due to the cumulative effect of the error statistic. Based on the scene classification results and statistical data, the following judgment conditions are determined: in This represents the GNSS measurement variance that can be estimated based on changes in the scene. and These represent the minimum and maximum values of the preset GNSS measurement variance, respectively.
8. A scene-classification-based pedestrian GNSS / INS adaptive integrated navigation system, characterized in that, include: The preprocessing module is used to collect raw GNSS data and preprocess the raw GNSS data to obtain preprocessed GNSS data. The classification module is used to classify the raw GNSS data using a GRU-based scene classification model to obtain scene classification results; The positioning module is used to acquire IMU data, combine the preprocessed GNSS data and the scene classification results, and obtain the integrated navigation result based on the improved adaptive Kalman filter algorithm.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the scene classification-based pedestrian GNSS / INS adaptive integrated navigation method as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the scene classification-based pedestrian GNSS / INS adaptive integrated navigation method as described in any one of claims 1 to 7.