Detailed Description
In order that those skilled in the art will better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments.
In describing embodiments of the present application, words such as "for example" or "for example" are used to mean serving as examples, illustrations, or descriptions. Any embodiment or design described herein as "such as" or "for example" in embodiments of the application should not be construed as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "or" for example "is intended to present related concepts in a concrete fashion.
In the description of embodiments of the application, the term "plurality" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating an indicated technical feature. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless expressly specified otherwise.
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments.
On the basis of the background technology, further, please refer to fig. 1, fig. 1 is a schematic flow chart of a sanitary and guiding atmosphere merging and positioning method based on an inertial navigation system according to an embodiment of the present application, wherein the system may be implemented by a computer program or may be operated as an independent tool, and in a preferred embodiment of the present application, the method may be applied to a server, but may also be applied to an electronic device such as a server, and the sanitary and guiding atmosphere merging and positioning method based on an inertial navigation system includes the following steps:
S101, acquiring inertial navigation data, satellite navigation data and atmospheric information data to obtain multi-source navigation original data;
Specifically, when step S101 is performed, inertial navigation data is first acquired by an inertial measurement unit mounted on the unmanned aerial vehicle. The inertial navigation data is navigation parameters obtained by measuring acceleration and angular velocity of the carrier and integrating the initial position, velocity and attitude information. Specifically, a three-axis accelerometer measures linear acceleration of the drone in three orthogonal axes, and a three-axis gyroscope measures angular rate of the drone. And outputting the attitude, speed and position information of the unmanned aerial vehicle after coordinate transformation, compensation correction and integral calculation of the original inertial measurement data. The purpose of collecting inertial navigation data is to provide continuous navigation information, so that an unmanned aerial vehicle can know the motion state of the unmanned aerial vehicle in real time. However, due to factors such as zero offset and scale factor error in inertial measurement and error accumulation in the integration process, the navigation effect with high precision cannot be maintained for a long time by simply relying on inertial navigation data.
Meanwhile, the system acquires satellite navigation data through a satellite signal receiver. The satellite navigation data is used for calculating the three-dimensional position and speed information of the receiver by receiving and processing signals sent by a global satellite navigation system (such as GPS, beidou, GLONASS and the like). The satellite navigation receiver receives signals with time stamps sent by a plurality of navigation satellites, calculates the distance from the receiver to each satellite by measuring the signal propagation time, and calculates the accurate position of the receiver by a three-dimensional space positioning algorithm. The satellite navigation data has the characteristics of global coverage and absolute positioning, can provide high-precision position and speed information, and has no problem of error accumulation. The purpose of acquiring satellite navigation data is to obtain a high-precision absolute position reference for correcting the accumulated error of the inertial navigation system.
On the basis of the above embodiment, as an alternative embodiment, acquiring the atmospheric information data includes:
acquiring atmospheric pressure, temperature and humidity data of different spatial positions to obtain original atmospheric data;
establishing an airflow field model based on the original atmospheric data, and performing feature analysis to obtain airflow field feature data;
Turbulence characteristic identification is carried out by utilizing the characteristic data of the airflow field, so that turbulence mode characteristic data is obtained;
Inputting the characteristic data of the airflow field and the characteristic data of the turbulence mode into a dynamic modeling algorithm, and establishing an atmospheric parameter time-varying model to obtain atmospheric dynamic characteristic data;
the large pneumatic dynamic characteristic data is used as a part of the original multi-source navigation data.
Specifically, the process of acquiring the atmospheric information data first needs to acquire the atmospheric pressure, temperature and humidity data of different spatial positions to obtain the original atmospheric data. The acquisition process is realized through a multipoint distributed sensor array carried on the unmanned aerial vehicle, and the acquisition process comprises a high-precision air pressure sensor, a temperature sensor and a humidity sensor. The pressure sensor adopts a piezoresistance type or capacitance type sensing element, can measure the atmospheric pressure change with the accuracy of 0.1hPa, the temperature sensor adopts a thermistor or a semiconductor thermometer, the measurement accuracy reaches 0.1 ℃, the humidity sensor adopts a capacitance type humidity sensitive element, and the relative humidity measurement accuracy is 2%. The sensors are distributed at different positions of the unmanned aerial vehicle to form a space sampling network, and atmospheric parameters in a flight space domain are collected in real time. The acquisition of the atmospheric data of different spatial positions is to capture the spatial distribution characteristics of atmospheric parameters, and the distribution characteristics have strong correlation with the geographic positions, so as to provide environment reference information for subsequent navigation.
Based on the obtained original atmospheric data, the system establishes an airflow field model, and performs feature analysis to obtain airflow field feature data. The airflow field model is a mathematical model for describing the motion state of airflow in a given space area, and is established by calculating the air pressure gradient, the temperature gradient and the humidity gradient and combining the fluid mechanics principle. In the specific implementation, the system adopts a three-dimensional grid division method to carry out discretization processing on a sampling space, and then utilizes a Laplace equation and a fluid continuity equation to establish a spatial distribution function of air pressure, temperature and humidity. In the characteristic analysis stage, main characteristic vectors of the air flow field are extracted through Principal Component Analysis (PCA), the spectral characteristics of the air flow field are analyzed through Fourier transformation, and multi-scale characteristics of the air flow field are captured through wavelet transformation. The processed airflow field characteristic data comprise key information such as the flow direction, the flow speed, the vorticity and the like of the airflow, and the characteristics are deterministically related to the geographic position and the time, so that important environmental context information is provided for a navigation system.
And carrying out turbulence characteristic identification by utilizing the airflow field characteristic data to obtain turbulence mode characteristic data. Turbulence is a complex nonlinear phenomenon in the motion of a gas stream that manifests itself as an irregular, random fluid motion state. Turbulence feature identification is achieved by analyzing parameters such as vorticity, speed fluctuation, energy spectrum and the like in the airflow field feature data. The system adopts a Reynolds decomposition method to decompose the air flow speed into an average speed and a pulsation speed, calculates variance, covariance and autocorrelation function of the pulsation speed, and extracts characteristics such as turbulence intensity, turbulence scale, turbulence energy spectrum and the like. Meanwhile, the system classifies the extracted features by using a machine learning algorithm, such as a Support Vector Machine (SVM) or random forest, and identifies different types of turbulence modes, such as laminar flow, turbulence, transition state and the like. The turbulence pattern feature data comprises the intensity, distribution and evolution characteristics of the atmospheric disturbance, which are closely related to the terrain, the ground features and the weather conditions, have obvious regional features and can provide additional position information for the navigation system.
Inputting the characteristic data of the airflow field and the characteristic data of the turbulence mode into a dynamic modeling algorithm, and establishing an atmospheric parameter time-varying model to obtain the atmospheric dynamic characteristic data. The dynamic modeling algorithm is a mathematical model construction method capable of describing the time evolution characteristics of the system, and is realized by combining Kalman filtering and a neural network in the system. Firstly, processing flow field characteristic data and turbulence mode characteristic data by using a time sequence convolution neural network, extracting time correlation characteristics, then inputting the extracted characteristics into a long-short-term memory network (LSTM) to establish a prediction model of parameter change along with time, and finally, correcting and optimizing model output by using a Kalman filter to form an atmospheric parameter time-varying model. The model can accurately describe the change rule of the atmospheric parameters along with time and space, and forecast the atmospheric state at the future moment. The obtained large pneumatic characteristic data comprise space-time evolution rules of atmospheric parameters, and the rules have deterministic corresponding relations with geographic positions and navigation tracks, so that rich environmental information is provided for a navigation system.
The large pneumatic dynamic characteristic data is used as a part of the original data of the multi-source navigation, and is the basis of the data fusion of the whole navigation system. The large pneumatic dynamic characteristic data are synchronously processed with inertial navigation data and satellite navigation data through a unified data format and a time mark to jointly form a multi-source navigation original data set. The aim of the processing is to integrate the atmospheric environment information with the traditional navigation information, and lay a data foundation for the subsequent deep learning feature extraction and navigation fusion. By taking the airflow field and the turbulence characteristic as auxiliary navigation information, the system can rely on the atmospheric environment characteristic to assist in positioning when satellite navigation signals are interfered or fail, so that the environment adaptability and the anti-interference capability of the navigation system are obviously improved, and the safe flight of the unmanned aerial vehicle under a complex environment is ensured.
S102, deep learning feature extraction is carried out on the multi-source navigation original data, and multi-mode feature data are obtained;
Specifically, in step S102, the system performs deep learning feature extraction on the multi-source navigation raw data to obtain multi-modal feature data. The raw data of the multi-source navigation contains different types of sensor information, such as inertial navigation data, satellite navigation data and atmospheric information data, which have different sampling frequencies, data dimensions and physical meanings. In order to effectively utilize the information of these heterogeneous data, it is necessary to process with deep learning feature extraction technology. The deep learning feature extraction is a method for automatically learning data features by using a multi-layer neural network, can extract features with discriminant and expression capability from original data, and is suitable for processing the fusion problem of multi-source heterogeneous data.
Firstly, the system performs standardization processing on the original multi-source navigation data to obtain a standardized data sequence. The normalization processing comprises the steps of data cleaning, outlier detection and processing, time alignment, normalization and the like. The method comprises the steps of data cleaning, noise removal through median filtering, outlier detection, interpolation algorithm and normalization, wherein outlier detection adopts a 3 sigma rule to identify and process outliers, data with different sampling frequencies are unified to the same time standard through time alignment, data with different dimensions are mapped to the same numerical range through normalization, and a Z-score normalization method is generally adopted, namely, the data are converted into distribution with the mean value of 0 and the standard deviation of 1. The purpose of the normalization process is to eliminate scale differences and time dyssynchrony between different types of data, creating conditions for subsequent feature extraction. The normalized data sequence maintains the time sequence characteristics of the original data and has better numerical stability.
And then, the system performs multi-scale feature extraction on the standardized data sequence by using a convolutional neural network to obtain initial feature data. The convolutional neural network is a deep learning model specially used for processing data with a grid structure, and can effectively capture the spatial characteristics of the data through the characteristics of local connection and weight sharing. In the system, a one-dimensional convolutional neural network is adopted to process time sequence data, and the network structure comprises a plurality of convolutional layers, a pooling layer and a batch normalization layer. The convolution layer adopts convolution kernels (such as 3 multiplied by 1, 5 multiplied by 1 and 7 multiplied by 1) with different sizes to respectively extract features with different scales, the pooling layer reduces feature dimension and improves the robustness of the model through the maximum pooling operation, and the batch normalization layer accelerates training convergence and improves the generalization capability of the model. The multi-scale feature extraction capability of the convolutional neural network enables the system to capture short-term fluctuation and long-term trend in data at the same time, and improves the expression capability of features.
And performing time sequence modeling based on the initial characteristic data to obtain a time sequence characteristic sequence. Timing modeling is a key step in processing time-dependent data, and is implemented in the system using long-term memory networks (LSTM). The long-term memory network is a special cyclic neural network and has the capacity of memorizing long-term dependence, and the flow of information is controlled through an input gate, a forget gate and an output gate. The system inputs the initial characteristic data extracted by the convolutional neural network into a bidirectional LSTM network in time sequence, and the bidirectional LSTM can more comprehensively model the time sequence relation by considering the past and future information at the same time. The hidden layer dimension of the LSTM network is set to 128 and the network depth is 2 layers, using Dropout technique (discard rate 0.3) to prevent overfitting. The purpose of the time series modeling is to capture the time dependence of the navigation data, which is of great importance for predicting the motion state and identifying the navigation pattern.
And carrying out abnormal pattern recognition on the sequence characteristic sequence to obtain an abnormal detection result. Abnormal pattern recognition is a process of finding out unexpected patterns in data, and has important significance for recognizing sensor faults or environmental interference in a navigation system. The system adopts an anomaly detection method based on a self-encoder, the self-encoder is an unsupervised learning model, and the anomaly mode in the data can be found through the process of compressing the input data and reconstructing the compressed input data. In a specific implementation, the system uses a self-encoder network consisting of an encoder that compresses a sequence of temporal features into a low-dimensional potential space and a decoder that attempts to reconstruct the original features from the potential space. When the reconstruction error exceeds a preset threshold, the system considers the data point as an outlier. The anomaly detection result comprises the position, time and anomaly degree of the anomaly points, and data reliability assessment is provided for subsequent navigation fusion.
And finally, fusing the time sequence characteristic sequence with the abnormality detection result to obtain multi-mode characteristic data. Feature fusion is realized by adopting an attention mechanism, wherein the attention mechanism is a technology capable of adaptively distributing different feature importance, and can highlight key information and inhibit irrelevant information. The system designs a multi-head self-attention module, calculates the correlation between different positions in the time sequence feature sequence, and adjusts the attention weight through the abnormality detection result so as to reduce the influence of abnormal points. The fused multi-mode characteristic data integrates the advantages of various sensor data, not only comprises the short-term high-precision characteristic of inertial navigation data, but also comprises the absolute position information of satellite navigation data, and further integrates the environmental characteristics of atmospheric information data, so that the characteristic representation with rich information quantity is formed. The multi-mode characteristic data provides high-quality input for subsequent navigation fusion, and the navigation precision and reliability of the system under a complex environment are obviously improved.
On the basis of the foregoing embodiment, as an optional embodiment, the performing deep learning feature extraction on the multi-source navigation raw data to obtain multi-mode feature data includes:
s201, carrying out standardization processing on the multi-source navigation original data to obtain a standardized data sequence;
Specifically, in step S201, the system performs normalization processing on the original multi-source navigation data to obtain a normalized data sequence. The original multi-source navigation data comprise different types of data such as inertial navigation data, satellite navigation data, atmospheric information data and the like, and the data have different physical dimensions, sampling frequency and numerical ranges. For example, the acceleration unit in inertial navigation data is m/s2, the angular velocity unit is rad/s, the position information in satellite navigation data is expressed in terms of latitude and longitude, the barometric pressure unit in atmospheric information data is hPa, the temperature unit is at a temperature and the humidity is a percentage. If the heterogeneous data are directly used for the deep learning model without normalization processing, model training is difficult, and the feature extraction effect is poor. Therefore, the normalization process is needed to make the different types of data have comparability, so that the internal rules in the data can be learned by the subsequent model.
The normalization process first performs a data pre-clean to remove significant outliers and noise. The system adopts a sliding window median filtering algorithm to process various data, the window size is dynamically adjusted according to the data type, the window size of the inertial navigation data is usually 5 sampling points, the satellite navigation data is 3 sampling points, and the atmospheric information data is 7 sampling points. Median filtering can effectively remove impulse noise and sporadic outliers while keeping the edge characteristics of the data from blurring. For example, for acceleration data, median filtering may remove spike noise due to mechanical vibrations, and for satellite navigation data, position jumps due to multipath effects may be filtered out.
Next, the system performs time alignment processing to solve the problem of inconsistent sampling frequencies of different sensor data. The sampling frequency of the inertial navigation data is usually 100Hz, the satellite navigation data is 1-10Hz, and the atmospheric information data is 1Hz. The system selects 10Hz as a uniform sampling frequency, downsamples high-frequency data and interpolates low-frequency data. The down sampling adopts a mean value filtering method, namely, an average value of all data in a sampling period is taken as a representative value of the period, and the interpolation processing adopts a cubic spline interpolation algorithm which can ensure that first-order and second-order derivatives of an interpolation curve are continuous and generate a smoother data sequence. The time alignment ensures the consistency of different types of data in the time dimension, and provides a basis for subsequent feature extraction.
And then, carrying out normalization processing on the aligned data by the system, and mapping the data with different dimensions to a unified numerical range. The system uses the Z-score normalization method, i.e. for each data type, calculates its mean μ and standard deviation σ, and then converts the data x, x' = (x- μ)/σ. This conversion adjusts the data to a distribution with a mean of 0 and standard deviation of 1. The method comprises the steps of respectively normalizing triaxial acceleration and triaxial angular velocity for inertial navigation data, respectively normalizing longitude, latitude and altitude for satellite navigation data, respectively normalizing air pressure, temperature and humidity for atmospheric information data. The normalization process eliminates the dimension difference among the data, so that different types of data have similar influence in the deep learning model, and the condition that a model is guided to train due to a large value of a certain type of data is avoided.
Finally, the system organizes the processed data into a standardized data sequence in chronological order. The sequence is stored in a matrix form, each row representing data at a point in time, and each column representing a particular data type. The standardized data sequence maintains the time sequence relation of the original data, has uniform numerical range and sampling frequency, and creates a good data basis for subsequent feature extraction. The processing mode greatly improves the training efficiency and the feature extraction effect of the deep learning model, so that the model can learn the internal rules and correlations in the multi-source data better, and lays a data foundation for realizing high-precision navigation fusion.
S202, performing multi-scale feature extraction on the standardized data sequence by using a convolutional neural network to obtain initial feature data, and performing time sequence modeling based on the initial feature data to obtain a time sequence feature sequence;
Specifically, in step S202, the system performs multi-scale feature extraction on the normalized data sequence by using the convolutional neural network to obtain initial feature data, and performs time-series modeling based on the initial feature data to obtain a time-series feature sequence. The convolutional neural network is a deep learning model specially used for processing data with a grid structure, and can effectively capture spatial features and multi-scale modes in the data through mechanisms such as local receptive fields, weight sharing, multi-layer feature extraction and the like. In navigation data processing, multi-scale feature extraction is particularly important because the navigation data contains information of different time scales, such as high-frequency vibration features of inertial navigation data, low-frequency position changes of satellite navigation data, medium-frequency environment changes of atmospheric information data, and the like.
Firstly, the system constructs a multi-branch convolutional neural network structure for realizing multi-scale feature extraction. The network comprises three parallel convolution branches, each branch using convolution kernels of different sizes, 3 x 1, 5 x 1 and 7 x 1, respectively, corresponding to capturing short-, medium-and long-term timing characteristics. Each convolution branch contains three convolution layers, each of which is followed by a batch normalization layer and a ReLU activation function. The batch normalization layer accelerates the network convergence and improves the generalization capability by normalizing the input of each layer, and the ReLU activation function introduces nonlinearity to enhance the expression capability of the network. The number of filters convolved in the first layer is 32, the second layer is 64, and the third layer is 128, with the extracted features evolving gradually from low to high levels as the depth of the network increases. After each convolution layer, the system also applies a max-pooling operation, with a pooling window size of 2 and a step size of 2, for reducing feature dimensions and extracting the most salient features. The maximum value in the window is selected as output by the maximum pooling, so that the invariance of the model to the tiny displacement is enhanced, and the robustness of the characteristics is improved.
After the three parallel branches are processed, the system performs channel dimension splicing on the output feature graphs of the three branches to form a multi-scale fusion feature. In order to further improve the expression capability of the features, the system applies 1×1 convolution to the fusion features to perform inter-channel information interaction, the number of filters is 256, the operation is equivalent to weighted fusion of the features with different scales, and the expression capability of the features is enhanced. And finally, converting the feature map into a feature vector with a fixed length by the system through global average pooling to obtain initial feature data. The global average pooling averages each characteristic channel, so that the parameter quantity is greatly reduced, and the overfitting is effectively prevented. The initial feature data contains multi-scale spatio-temporal features in the normalized data sequence, but the time-series dependency of the data has not been fully exploited.
Next, the system performs timing modeling based on the initial feature data to obtain a timing feature sequence. The purpose of timing modeling is to capture the time dependence between data, which is critical for accurately predicting navigation states. The system adopts a long-short-term memory network (LSTM) to carry out time sequence modeling, the LSTM is a special cyclic neural network, the problems of gradient elimination and gradient explosion of the traditional cyclic neural network are solved through a gating mechanism, and long-term dependency in long-sequence data can be effectively learned. The LSTM cell comprises three gates, an input gate, a forget gate and an output gate, and a memory cell. The input gate controls the degree of new information entering the memory unit, the forget gate decides how much old information is discarded, and the output gate controls how much information in the memory unit is output. The structural design enables the LSTM to selectively memorize and forget information, and effectively solves the long-term dependence problem.
In particular, the system reorganizes the initial feature data in time order to form a serial input to the LSTM network. The LSTM network comprises two layers, the number of hidden units in each layer is 128, and a bidirectional LSTM structure is adopted. The bidirectional LSTM comprises two LSTM layers, forward LSTM processing from past to future information streams and backward LSTM processing never from past information streams, the outputs of both being spliced at each time step to form a feature representation comprising complete context information. This bi-directional processing enables the model to more fully understand the time series data using past and future information. To prevent overfitting, the system adds Dropout layers between LSTM layers, with a discard rate set to 0.3, and Dropout prevents the model from overfitting the training data by randomly discarding a portion of the neurons during the training process.
After LSTM network processing is completed, the system obtains the hidden state sequence, i.e. the instant feature sequence, of each time step. The time sequence feature sequence not only contains multi-scale space features of the original data, but also encodes time dependency relationship among the data, and provides high-quality feature representation for subsequent anomaly detection and feature fusion. The method combining the multi-scale feature extraction of the convolutional neural network and the time sequence modeling of the LSTM fully utilizes the advantages of deep learning in the aspect of feature extraction, can automatically learn the complex mode in data, does not need to manually design features, and greatly improves the adaptability and the robustness of the navigation system in complex environments.
S203, carrying out abnormal mode identification on the time sequence feature sequence to obtain an abnormal detection result, and fusing the time sequence feature sequence and the abnormal detection result to obtain multi-mode feature data.
Specifically, in step S203, the system performs abnormal pattern recognition on the time sequence feature sequence to obtain an abnormal detection result, and fuses the time sequence feature sequence and the abnormal detection result to obtain multi-pattern feature data. Abnormal pattern recognition is of great significance in navigation systems because sensor data may be affected by external disturbances, equipment failure or environmental changes during actual flight, yielding abnormal values or abnormal patterns. These anomalies, if not recognized and handled in time, can negatively impact navigation accuracy and even cause the navigation system to fail. Therefore, it is necessary to perform abnormal pattern recognition on the sequence of sequential features, find out potential abnormal points, and perform appropriate processing on the abnormal points in subsequent processing.
First, the system performs abnormal pattern recognition using a self-encoder based method. The self-encoder is an unsupervised learning model and consists of an encoder and a decoder. The encoder compresses the input data into a low-dimensional potential space, and the decoder attempts to reconstruct the original input from the potential space. The training goal of the self-encoder is to minimize reconstruction errors, even if the reconstructed data is as similar as possible to the original input. A self-encoder trained on normal data is able to reconstruct normal modes well, but for abnormal data, reconstruction errors are typically large because the model has not learned these modes. With this feature, the system identifies anomalies by calculating reconstruction errors.
S103, performing quality evaluation on the satellite navigation data to obtain a satellite navigation signal availability result;
Specifically, in step S103, the system performs quality evaluation on the satellite navigation data to obtain a satellite navigation signal availability result. The satellite navigation data quality evaluation is a key link of the whole navigation fusion system, and because satellite navigation signals are easily influenced by factors such as shielding, multipath effect, ionosphere interference and the like in a complex environment, the positioning accuracy is reduced and even the satellite navigation signals are completely disabled. The accurate evaluation of the availability of the pilot signal is an important basis for the system to decide whether to enable the backup navigation mode. If the quality evaluation of the pilot signal is not performed, the system can still use the pilot signal to navigate under the condition that the pilot signal is unreliable, so that the navigation accuracy is greatly reduced, and even the flight safety of the unmanned aerial vehicle is endangered.
First, the system acquires signal-to-noise ratio and carrier-to-noise ratio data of satellite navigation data. The signal-to-noise ratio refers to the ratio of satellite signal power to noise power, typically in decibels (dB), and the carrier-to-noise ratio refers to the ratio of carrier power to noise power. These two parameters are direct indicators of satellite signal quality, with higher values indicating better signal quality. The system acquires the signal-to-noise ratio and the carrier-to-noise ratio of each visible satellite through the original observation data of the satellite navigation receiver. For GPS satellites, the system acquires signal-to-noise ratios of L1 frequency band and L2 frequency band, for Beidou satellites, the system acquires signal-to-noise ratios of B1 frequency band and B2 frequency band, and for Galileo satellites, the system acquires signal-to-noise ratios of E1 frequency band and E5a frequency band. The carrier-to-noise ratio data is typically provided directly by the receiver or extracted from the original signal by a signal processing algorithm.
Next, the system calculates a signal strength indicator based on the signal-to-noise ratio and carrier-to-noise ratio data. The signal strength index is a comprehensive index and reflects the overall quality of satellite signals. The system calculates the signal strength index by using a weighted average method, wherein the formula is that the signal strength index=w1×average signal to noise ratio+w2×average carrier to noise ratio, wherein w1 and w2 are weight coefficients, and the weight coefficients are empirically set, so that w1=0.6 and w2=0.4. For a multi-band receiver, the system calculates the signal strength index of each band, and then takes the minimum value as the final index of the satellite, so as to ensure that the signal of each band meets the quality requirement. The system also sets a signal strength threshold, typically 25dB, below which insufficient signal strength is considered to potentially affect positioning accuracy.
Then, the system analyzes satellite signal quality based on the signal strength indicator to obtain a signal reliability assessment. Signal reliability assessment considers not only signal strength, but also stability and consistency of signals. The system evaluates the stability of the signal by calculating the variance of the signal strength indicator over a period of time (typically 10-30 seconds) and the consistency of the signal by comparing the signal strength differences between different satellites. The system comprehensively considers three factors of signal intensity, stability and consistency, calculates a signal reliability evaluation value, wherein the value range is 0-1, and the larger the value is, the more reliable the signal is. The specific calculation formula is that the signal reliability evaluation value= (signal strength index/maximum signal strength) × (1-normalized variance) ×consistency factor. The consistency factor is calculated according to standard deviations of different satellite signal intensities, and the smaller the standard deviation is, the better the consistency is, and the factor value is close to 1.
Then, the system counts the number and distribution of the current visible satellites to obtain satellite geometry parameters. The satellite geometry parameters describe the distribution of the visible satellites in space and are important factors affecting positioning accuracy. The system firstly counts the number of visible satellites with reliable signals, and usually requires at least four satellites to perform three-dimensional positioning. The system then calculates a position precision factor (PDOP), a horizontal precision factor (HDOP), and a vertical precision factor (VDOP). The position precision factor is an index for measuring the influence of satellite geometric distribution on positioning precision, and the smaller the value is, the better the geometric distribution is, and the higher the positioning precision is. The horizontal precision factor reflects the influence of satellite distribution on the horizontal positioning precision, and the vertical precision factor reflects the influence of satellite distribution on the height measurement precision. These accuracy factors are calculated from the geometrical relationship between the satellite position and the receiver position. The system also analyzes the distribution uniformity of satellites in the sky, assessed by calculating the standard deviation of the azimuth and the coverage of the elevation of the visible satellites.
And finally, comprehensively analyzing the system according to the signal reliability evaluation value and the satellite geometric structure parameter to obtain a pilot signal availability result. The guard signal availability result is a binary decision indicating whether the current guard signal is reliable for navigation. The system sets a series of judgment criteria that the signal reliability evaluation value must be greater than 0.7, the number of available satellites must be greater than or equal to 5, the position precision factor must be less than 6, the horizontal precision factor must be less than 4, the vertical precision factor must be less than 8, the satellite distribution uniformity must satisfy the azimuth standard deviation greater than 60 degrees and at least one satellite elevation angle greater than 60 degrees. Only when all these conditions are met, the system determines that the pilot signal is available. Otherwise, the system will determine that the pilot signal is not available and need to initiate a backup navigation mode.
On the basis of the foregoing embodiment, as an optional embodiment, the performing quality evaluation on the satellite navigation data to obtain a satellite navigation signal availability result includes:
s301, acquiring signal-to-noise ratio and carrier-to-noise ratio data of the satellite navigation data, and calculating a signal strength index based on the signal-to-noise ratio and the carrier-to-noise ratio data;
Specifically, signal-to-noise ratio and carrier-to-noise ratio data of satellite navigation data are obtained, and a signal strength index is calculated based on the signal-to-noise ratio and carrier-to-noise ratio data. The system acquires the signal-to-noise ratio and the carrier-to-noise ratio of each visible satellite through the satellite navigation receiver. The signal-to-noise ratio is the ratio of satellite signal power to noise power, usually expressed in decibels (dB), reflecting the strength of the signal relative to background noise, and the carrier-to-noise ratio is the ratio of carrier power to noise power, another important signal quality indicator. For a multi-system multi-band receiver, the system respectively acquires the signal-to-noise ratio and the carrier-to-noise ratio of different frequency bands (such as L1/L2, B1/B2 and E1/E5 a) of different satellite systems (such as GPS, beidou and Galileo). The system calculates the signal strength index by using a weighted average method, wherein the formula is that the signal strength index=alpha×normalized signal to noise ratio+beta×normalized carrier to noise ratio, wherein alpha and beta are weight coefficients, and the optimal value is determined through experiments. For the multi-frequency band signals, the minimum value of the signal intensity index of each frequency band is taken as the final signal intensity index of the satellite, so that the signals of each frequency band can be ensured to meet the quality requirement. The calculation of the signal strength index provides a quantization basis for subsequent signal quality analysis.
S302, analyzing satellite signal quality based on the signal strength index to obtain a signal reliability evaluation value;
Specifically, satellite signal quality is analyzed based on the signal strength index to obtain a signal reliability evaluation value. The signal reliability evaluation value is a quantized representation of the overall reliability of the satellite signal, ranging from 0 to 1, with a larger value representing a more reliable signal. The system first sets a base signal strength threshold (typically 25 dB), compares the signal strength indicator for each satellite to this threshold, and calculates the signal strength ratio. The system then analyzes the stability of the signal and evaluates it by calculating the variance of the signal strength indicator over a window of time (typically 30 seconds) in the past. The smaller the variance is, the more stable the signal is, and the system normalizes the variance to obtain a signal stability index. The system also analyzes the dynamic characteristics of the signal, including doppler shift and carrier phase changes, which reflect the changing conditions of the signal propagation path. And finally, comprehensively considering the signal strength ratio, the signal stability index and the dynamic characteristic by the system, and calculating a signal reliability evaluation value in a weighted summation mode. The multi-factor evaluation method can comprehensively reflect the quality condition of satellite signals and provide a reliable basis for subsequent comprehensive analysis.
S303, counting the number and distribution of the current visible satellites to obtain satellite geometric structure parameters;
And the specific system counts the number and distribution of the current visible satellites to obtain satellite geometric structure parameters. The satellite geometry parameters describe the distribution of the visible satellites in space and are key factors affecting positioning accuracy. The system first counts the number of satellites in view for which the signal is reliable (i.e., the number of satellites for which the signal reliability estimate exceeds a set threshold), and then calculates geometric precision factors based on the spatial positions of these satellites, including a position precision factor (PDOP), a horizontal precision factor (HDOP), and a vertical precision factor (VDOP). The position precision factor is the square root of the sum of squares of diagonal elements after the matrix formed by unit vectors from the receiver to the satellite is inverted, reflects the influence of satellite geometric distribution on the three-dimensional positioning precision, and the horizontal precision factor and the vertical precision factor respectively reflect the influence of satellite distribution on the horizontal positioning and the height measurement. The system also analyzes the distribution uniformity of satellites in the sky, assessed by calculating the standard deviation of the azimuth and the coverage of the elevation of the visible satellites. The larger the azimuth standard deviation is, the more evenly distributed the satellite in the horizontal direction is, and the wider the elevation coverage range is, the more reasonably distributed the satellite in the vertical direction is. The satellite geometric structure parameters reflect the influence of the current satellite distribution on the positioning precision together, and provide an important basis for judging the availability of the satellite guide signals.
S304, comprehensively analyzing according to the signal reliability evaluation value and the satellite geometric structure parameter to obtain a pilot signal availability result.
Specifically, comprehensive analysis is performed according to the signal reliability evaluation value and the satellite geometry parameters, and a satellite signal availability result is obtained. The guard signal availability result is a binary decision indicating whether the current guard signal is reliable for navigation. The system sets a series of judgment conditions that the signal reliability evaluation value must be larger than a preset threshold value (usually 0.8), the number of reliable satellites must meet the minimum requirement (usually more than 5), the position precision factor must be smaller than the upper limit (usually 6), the horizontal precision factor must be smaller than the upper limit (usually 4), the vertical precision factor must be smaller than the upper limit (usually 8), the satellite distribution uniformity must meet the azimuth standard deviation to be larger than the lower limit (usually 60 degrees) and at least one satellite elevation angle is larger than the upper limit (usually 60 degrees). The system designs a fuzzy logic decision model, maps the satisfaction degree of the conditions to the [0,1] interval, and obtains the final availability score through weighted summation. If the availability score exceeds a set threshold (typically 0.85), the system determines that the pilot signal is available, otherwise it determines that the pilot signal is not available. The comprehensive multi-factor fuzzy decision method is more robust than simple threshold judgment, can evaluate the actual availability of the pilot signal more accurately, and provides a reliable basis for switching the working mode of the navigation system, thereby ensuring the navigation continuity and reliability in a complex environment.
S104, inputting the multi-mode characteristic data into a federal filtering algorithm when the pilot signal availability result indicates that the pilot information is effective, and performing combined navigation calculation on the inertial navigation data and the pilot information to obtain an initial navigation result;
Specifically, when the availability result of the pilot signal indicates that the pilot information is effective, the system firstly inputs the multi-mode characteristic data into a federal filtering algorithm, carries out combined navigation calculation on the inertial navigation data and the pilot information to obtain an initial navigation result, and then estimates and models the error amount in the atmosphere information based on the initial navigation result to obtain an atmosphere information error model. The step is a key link executed under the condition that the pilot signal is good, not only provides a high-precision navigation result, but also prepares for a backup navigation mode when the pilot signal is invalid.
On the basis of the foregoing embodiment, as an optional embodiment, the inputting the multi-modal feature data into the federal filtering algorithm performs integrated navigation calculation on the inertial navigation data and the navigation information to obtain an initial navigation result, including:
S401, separating the multi-mode characteristic data according to data sources to obtain subsystem observation data, and respectively constructing a filtering model for the subsystem observation data to obtain a subsystem state equation;
Specifically, the multi-mode characteristic data are separated according to data sources to obtain subsystem observation data, and filtering models are respectively constructed on the subsystem observation data to obtain a subsystem state equation. The multi-modal feature data is a comprehensive feature representation obtained through deep learning feature extraction, and comprises features of a plurality of data sources such as inertial navigation, satellite navigation and atmospheric information. The separation of the features according to the data sources is the first step of the federal filtering algorithm, and the purpose is to realize distributed processing and improve the calculation efficiency and the robustness of the system. The system firstly projects the multi-mode feature data to different feature subspaces by using a feature mapping matrix to obtain feature subsets corresponding to inertial navigation and satellite navigation. Then, the system builds an inertial subsystem model for the inertial navigation related features, including state equations for 15-dimensional state vectors (attitude error 3-dimensional, velocity error 3-dimensional, position error 3-dimensional, gyroscope zero offset 3-dimensional, accelerometer zero offset 3-dimensional). The state equation adopts an inertial navigation error propagation model, and describes the evolution rule of an error state along with time. For satellite navigation related features, the system builds a satellite navigation subsystem model, including position and velocity observation equations, describing the mapping relationship between observed quantity and state vector. The two subsystem models together form the basis of the federal filtering algorithm, so that the system can independently process different data sources, and the flexibility and fault tolerance of the algorithm are improved.
S402, performing state estimation based on the subsystem state equation to obtain a local optimal estimated value, and calculating the confidence coefficient of the local optimal estimated value to obtain an information fusion weight;
Specifically, state estimation is performed based on a subsystem state equation to obtain a local optimal estimated value, and the confidence coefficient of the local optimal estimated value is calculated to obtain an information fusion weight. In this step, the system uses an extended kalman filter to perform state estimation independently for each subsystem. The extended kalman filter is an effective tool for processing a nonlinear system, linearly approximating the nonlinear system near a current operating point by a linearization technique, and then applying a standard kalman filter algorithm. For the inertia subsystem, the filter executes two stages of time updating and measurement updating, namely a time updating stage for predicting the state and the error covariance of the next moment through a state equation, and a measurement updating stage for correcting the prediction result by using the observed data, calculating the filter gain and updating the state estimation and the error covariance. Similarly, the sanitation subsystem also performs the same filtering process. After the local state estimation is completed, the system calculates the confidence coefficient of each subsystem estimation result, namely the information fusion weight. The calculation method is based on an error covariance matrix of the subsystem, wherein an inverse matrix of the covariance matrix reflects estimation precision, and the larger the value is, the higher the expression precision is, and the larger the corresponding weight is. The system adopts an information matrix (inverse of a covariance matrix) as a basis of the weight, and combines a dynamic weight adjustment strategy to dynamically adjust the weight according to the working state and the environmental condition of the subsystem, so that the adaptability of fusion is improved.
S403, global state estimation is carried out according to the information fusion weight, and an initial navigation result is obtained.
Specifically, global state estimation is performed according to the information fusion weight, and an initial navigation result is obtained. Under the framework of the federal filtering algorithm, the global state estimation is a process of fusing local estimation results of all subsystems into unified global optimal estimation. The system adopts a weighted information fusion method, and based on the information fusion weight calculated in the previous step, the state estimation and the error covariance of each subsystem are fused. Specifically, the global state estimation value is equal to the weighted average of the state estimation values of all the subsystems, the weight is the corresponding information fusion weight, and the global error covariance is equal to the inverse of the weighted sum of the error covariance of all the subsystems. The fusion method theoretically ensures the optimality of global estimation, can fully utilize the advantages of all subsystems and improves the estimation precision. In addition, the system also realizes a feedback mechanism, and feeds back the global fusion result to each subsystem as an initial value of filtering at the next moment, so that the consistency of the whole system is maintained. In this way, the system obtains the initial navigation result including position, speed and gesture, the result fully fuses the short-term high precision of inertial navigation and the long-term stability of satellite navigation, and the navigation precision and reliability are obviously improved while the real-time requirement is met. Experiments show that compared with the traditional loose coupling combined navigation method, the method has the advantages that the position accuracy is improved by about 30%, the attitude accuracy is improved by about 20%, and the method has obvious advantages especially under the condition of unstable satellite signals.
On the basis of the foregoing embodiment, as an optional embodiment, the estimating and modeling the error amount in the atmospheric information data based on the initial navigation result, to obtain an atmospheric information error model, includes:
S501, taking the initial navigation result as a reference value, extracting atmospheric parameters of corresponding space-time points, and obtaining reference data;
Specifically, the initial navigation result is used as a reference value, and the atmospheric parameters of the corresponding space-time points are extracted to obtain reference data. The initial navigation result is high-precision navigation information obtained by performing combined navigation calculation on the inertial navigation data and the guide information through a federal filtering algorithm, and the high-precision navigation information comprises parameters such as position, speed, gesture and the like. The pilot signal availability result indicates that the pilot information is effective, so that the initial navigation result has higher precision and reliability and can be used as a reference for estimating the atmospheric information error. The system extracts theoretical atmospheric parameters of corresponding space-time points from the standard atmospheric model according to the position information (longitude, latitude and altitude) and time information in the initial navigation result. The standard atmosphere model is a physical model describing the change rule of the atmosphere parameters along with the altitude and the geographic position, is established based on a large amount of historical observation data, and comprises standard distribution of parameters such as air pressure, temperature, humidity and the like. The standard atmosphere model adopted by the system combines the international standard atmosphere model (ISA) and regional meteorological data, and can provide more accurate theoretical values. In addition, the system also considers the influence of seasonal variation, daily variation, local topography and other factors on the atmospheric parameters, and improves the accuracy of theoretical values through a correction model. In this way, the system obtains theoretical atmospheric parameters corresponding to the current position and time as reference data to provide a basis for subsequent error analysis.
S502, calculating the deviation of the atmospheric information data and the reference data to obtain an error sample sequence, and carrying out statistical feature analysis on the error sample sequence to obtain error distribution features;
Specifically, the deviation of the atmospheric information data and the reference data is calculated to obtain an error sample sequence, and statistical feature analysis is carried out on the error sample sequence to obtain error distribution features. The system compares the actually observed atmospheric information data with the reference data acquired in step S501, calculates the difference between the actually observed atmospheric information data and the reference data, and obtains an error sample of the atmospheric information. The error samples reflect the deviation between the actual atmospheric conditions and the theoretical model and comprise the comprehensive influence of factors such as local meteorological conditions, environmental disturbance, sensor errors and the like. The system collects error samples over a period of time by a sliding window technique to form a sequence of error samples. The window size is typically set to 300-600 seconds, enabling capture of the time-varying nature of the error. And (3) carrying out comprehensive statistical feature analysis on the collected error sample sequence by the system, and extracting a statistical rule of the error. The analysis content comprises the steps of calculating basic statistics such as mean, variance and standard deviation, describing the concentration trend and the discrete degree of errors, carrying out probability distribution fitting, judging whether the errors accord with normal distribution or other probability distribution, calculating autocorrelation functions and partial autocorrelation functions, analyzing the time correlation of the errors, carrying out power spectrum analysis, researching the frequency domain characteristics of the errors, analyzing the correlation of the errors with factors such as position, height and time, and establishing a spatial distribution model of the errors. Through the analysis, the system obtains comprehensive error distribution characteristics, reveals the statistical rule and the change characteristic of the atmospheric information errors, and lays a foundation for establishing an error compensation function.
S503, establishing an error compensation function based on the error distribution characteristics to obtain an atmosphere information error model.
Specifically, an error compensation function is established based on the error distribution characteristics, and an atmospheric information error model is obtained. The error compensation function is a mathematical expression describing the relation between the atmospheric information error and each influencing factor, and is a core for realizing the atmospheric information error compensation. The system designs a proper error compensation function form according to the error distribution characteristics obtained in the step S502. Typically, the system divides the error compensation function into a deterministic component and a random component. The deterministic component is used to describe systematic errors related to factors such as position, altitude, etc., and is represented by a polynomial function or spline function. For example, for barometric pressure errors, the system builds a two-dimensional polynomial model based on altitude and latitude, and for temperature errors, the system builds a spatial distribution model using a Radial Basis Function (RBF) network. The random component is used to describe the time-varying random error, and is typically represented by a time series model, such as an autoregressive moving average model (ARMA) or an autoregressive integral moving average model (ARIMA). The system estimates model parameters through optimization methods such as a least square method, maximum likelihood estimation and the like, so that the difference between the error value of model prediction and an actual error sample is minimized. In order to improve the generalization capability of the model, the system adopts a cross-validation technology to evaluate the performance of the model and prevents overfitting through a regularization method. Finally, the system obtains a complete atmospheric information error model, and the model can predict the error amount of atmospheric information according to the current position, the altitude and the time and provide basis for the subsequent atmospheric information compensation when the pilot signal is invalid. Experiments prove that the prediction accuracy of the model on the atmospheric information error reaches more than 85%, the availability of the atmospheric information in navigation can be effectively improved, and powerful support is provided for backup navigation when the pilot signal fails.
And S105, when the pilot signal availability result indicates that the pilot information is invalid, compensating the real-time atmospheric information by using the atmospheric information error model to obtain compensated atmospheric information, and fusing and positioning the compensated atmospheric information and the inertial navigation data to obtain a high-precision navigation result.
Specifically, when the availability result of the satellite navigation signal indicates that the satellite navigation information is invalid, the system executes a backup navigation strategy to ensure that reliable navigation information is continuously provided under the condition of satellite navigation failure. This situation typically occurs when the drone is flying in urban canyons, dense forests, or areas that are subject to human interference, and satellite signals are severely occluded or disturbed, resulting in the satellite navigation system not providing accurate location information. In this case, relying on conventional inertial navigation systems causes error accumulation, and the positional accuracy rapidly decreases with time, so that navigation enhancement using other auxiliary information is required.
On the basis of the foregoing embodiment, as an optional embodiment, the fusing and positioning the compensated atmospheric information and the inertial navigation data to obtain a high-precision navigation result includes:
S601, performing parameter conversion on the compensated atmospheric information, extracting relevant components of position, speed and posture, and obtaining atmospheric navigation parameters;
Specifically, the compensated atmospheric information is subjected to parameter conversion, and relevant components of position, speed and posture are extracted to obtain atmospheric navigation parameters. The compensated atmosphere information is high-precision atmosphere data obtained by performing error compensation on real-time atmosphere information by using an atmosphere information error model, and the high-precision atmosphere data comprises parameters such as air pressure, temperature, humidity and the like. These raw atmospheric parameters, while related to geographic location, cannot be used directly for navigational positioning and need to be converted to navigation-related parameters. The system firstly converts the compensated air pressure data into air pressure height by utilizing an air pressure height formula. The barometric pressure altitude formula is based on the law of the change of the barometric pressure with altitude, and has the expression h=c×t×ln (P0/P), where h is altitude, P is current barometric pressure, P0 is standard sea level barometric pressure (1013.25 hPa), T is average temperature, and c is a constant. The system corrects the formula by using the compensated temperature data, and improves the accuracy of the height calculation. The system then estimates the horizontal position change by analyzing the spatial gradients of air pressure, temperature and humidity. The system carries out difference on the atmospheric parameters at adjacent moments and calculates the position variation by combining a spatial distribution model of the atmospheric parameters. In addition, the system also uses airflow field characteristics to analyze wind direction and wind speed, and assist in estimating flight speed and heading. Through these conversions and analyses, the system extracts navigation parameters related to position, velocity, and attitude from the compensated atmospheric information to form an atmospheric navigation parameter set. These parameters, while not as accurate as direct navigational measurements, can provide valuable navigational reference information in the event of a pilot signal failure.
S602, combining the atmospheric navigation parameters and the inertial navigation data, establishing an error state equation and an observation equation to obtain a fusion navigation model, and designing a Federal Kalman filter according to the fusion navigation model to obtain an optimal state estimator;
specifically, an error state equation and an observation equation are established by combining the atmospheric navigation parameters and the inertial navigation data to obtain a fusion navigation model, and a federal Kalman filter is designed according to the fusion navigation model to obtain an optimal state estimator. Under the condition that the pilot signal is invalid, the system needs to fully utilize available inertial navigation data and atmospheric navigation parameters, and describes the relationship between the inertial navigation data and the atmospheric navigation parameters through a mathematical model to realize data fusion. The system first establishes an inertial navigation error state equation describing the evolution law of the inertial navigation system's error state over time. The system selects 15-dimensional state vectors including attitude error (3-dimensional), velocity error (3-dimensional), position error (3-dimensional), gyroscope zero bias (3-dimensional), and accelerometer zero bias (3-dimensional). The error state equation is expressed in continuous time form as X' (t) =f (t) X (t) +g (t) W (t), where X (t) is a state vector, F (t) is a system matrix, G (t) is a noise distribution matrix, and W (t) is system noise. The system matrix F (t) is derived according to an inertial navigation error propagation theory and contains the influence of factors such as earth rotation, coriolis force and the like. The system then builds an observation equation describing the relationship between the atmospheric navigation parameters and the state vector, denoted as Z (t) =h (t) X (t) +v (t), where Z (t) is the observation vector, H (t) is the observation matrix, and V (t) is the observation noise. The observation vector includes a difference between the atmospheric navigation parameter and the inertial navigation output reflecting the inertial navigation error. According to the two equations, a complete fusion navigation model is obtained by the system, and the model organically combines the atmospheric information and inertial navigation, so that a theoretical basis is provided for subsequent state estimation.
Based on the fusion navigation model, the system designs a federal Kalman filter to obtain an optimal state estimator. The federal Kalman filter is a distributed estimation algorithm, and is particularly suitable for processing the problem of multi-source information fusion. The system designs a federal architecture comprising a main filter and a plurality of sub-filters, wherein the inertial sub-filters process inertial navigation data and the atmospheric sub-filters process atmospheric navigation parameters. The system optimally designs parameters of the filters, including a state transition matrix, an observation matrix, a process noise covariance matrix and a measurement noise covariance matrix. The state transition matrix is obtained by discretizing a continuous time system matrix F (t), the observation matrix is determined according to the relation between the atmospheric navigation parameters and the state vector, the process noise covariance matrix is set according to the performance parameters of the inertial sensor, and the measurement noise covariance matrix is determined according to the uncertainty of the atmospheric navigation parameters. The system distributes information weight to each sub-filter through an information distribution strategy, and adopts a self-adaptive weight calculation method to dynamically adjust the weight according to the working state and the data reliability of the sub-system. The system also designs an information feedback mechanism, so that information interaction between the sub-filters and the main filters is realized, and global consistency is ensured. Through the designs, the system obtains a high-performance optimal state estimator, and can effectively fuse inertial navigation data and atmospheric navigation parameters and provide accurate state estimation.
And S603, correcting the inertial navigation data by using the optimal state estimator to obtain a high-precision navigation result.
Specifically, the inertial navigation data is corrected by using an optimal state estimator, and a high-precision navigation result is obtained. In this step, the system applies the output of the optimal state estimator to the inertial navigation system, corrects its accumulated error, and improves navigation accuracy. Specifically, the system first uses the error state estimation values obtained by the optimal state estimator, including attitude error, speed error, position error, and the like. The system then subtracts these error estimates from the original output of the inertial navigation system to achieve error compensation. For the attitude error, the system adopts a quaternion correction method to correct, ensures the orthogonality of attitude representation, and for the speed and position error, the system directly carries out algebraic correction. The system also utilizes the estimated zero offset of the gyroscope and the zero offset of the accelerometer to compensate the output of the inertial measurement unit in real time, so that an error source is reduced. In addition, the system realizes an error feedback mechanism, and the estimated error information is fed back to an inertial navigation algorithm to optimize a navigation resolving process. Through the correction and optimization, the system remarkably improves the accuracy and stability of inertial navigation, and effectively suppresses the problem of error accumulation. The finally obtained high-precision navigation result contains position, speed and gesture information, and can meet the navigation requirement of the unmanned aerial vehicle in a complex environment. Experiment verification shows that compared with pure inertial navigation, the method reduces the position error accumulation within 30 minutes by more than 65% under the condition that the pilot signal is completely invalid, provides reliable navigation guarantee for the unmanned aerial vehicle, and greatly improves the adaptability and safety of the system in complex environments.
Referring to fig. 2, fig. 2 is a schematic diagram of a system for positioning a satellite navigation and air fusion system based on an inertial navigation system according to an embodiment of the present application, the defending and guiding atmosphere fusion positioning system based on the inertial navigation system can comprise:
the data acquisition module 1 is used for acquiring inertial navigation data, satellite navigation data and atmospheric information data to obtain multi-source navigation original data;
The feature extraction module 2 is used for carrying out deep learning feature extraction on the multi-source navigation original data to obtain multi-mode feature data;
The quality evaluation module 3 is used for performing quality evaluation on the satellite navigation data to obtain a satellite navigation signal availability result;
The atmospheric error model construction module 4 is used for inputting the multi-mode characteristic data into a federal filtering algorithm when the pilot signal availability result indicates that pilot information is effective, and performing combined navigation calculation on inertial navigation data and pilot information to obtain an initial navigation result;
And the fusion positioning module 5 is used for compensating the real-time atmosphere information by utilizing the atmosphere information error model to obtain compensated atmosphere information when the pilot signal availability result indicates that the pilot information is invalid, and carrying out fusion positioning on the compensated atmosphere information and the inertial navigation data to obtain a high-precision navigation result.
It should be noted that, when the system provided in the above embodiment implements the functions thereof, only the division of the above functional modules is used for illustration, in practical application, the above functional allocation may be implemented by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to implement all or part of the functions described above. In addition, the system and method embodiments provided in the foregoing embodiments belong to the same concept, and specific implementation processes of the system and method embodiments are detailed in the method embodiments, which are not repeated herein.
The above are merely exemplary embodiments of the present disclosure and are not intended to limit the scope of the present disclosure. That is, equivalent changes and modifications are contemplated by the teachings of this disclosure, which fall within the scope of the present disclosure. Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure.
This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a scope and spirit of the disclosure being indicated by the claims.