A GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning

Abstract

The present application relates to a GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning, including obtaining GNSS observation data, extracting and calculating GNSS data features, constructing a time-series GNSS feature vector of GNSS data features through the SFFS algorithm; constructing an LSTM neural network with time memory function based on the GNSS feature vector; training the LSTM neural network in combination with GNSS observation data and scene labels preset for navigation scenes to obtain an LSTM model; inversely calculating the residuals of pseudorange and carrier phase; fitting the functional relationship between the residuals and preset influencing factors to construct GNSS random models for different classification scenarios; constructing threshold models for different classification scenarios based on preset different classification scenarios and GNSS observation data, the threshold model including carrier-to-noise ratio, satellite altitude angle and ratio value. The present application has the effect of improving the availability, accuracy and continuity of GNSS positioning services in complex scenarios to meet the GNSS positioning needs in dynamic and complex environments.

Description

Technical Field

[0001] The present application relates to the technical field of GNSS, and in particular to a GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning. Background Art

[0002] The continuous development of multi-frequency and multi-mode fusion technologies such as RTD and RTK enables GNSS to provide users with centimeter-level high-precision dynamic positioning services in open environments. Through integration with vehicles, mobile phones, drones and other carriers, the application environment of GNSS has gradually changed from static, single, and simple to dynamic, diverse, and complex. This has put forward new requirements for the dynamic positioning performance of GNSS in complex environments.

[0003] The differential loss of GNSS signals caused by different scenarios makes the quality of GNSS observations between scenarios no longer consistent; however, the existing GNSS threshold model has multiple fixed parameters, which is obviously inconsistent with the actual quality distribution of GNSS observations; in terms of random model optimization, in order to characterize the differential quality of GNSS observations in different scenarios, existing technologies mostly reconstruct random models for interest sites. This technology optimizes random models for a single scenario. Such random models with fixed parameters are also unable to adapt to the GNSS positioning requirements in dynamic and complex environments; or the weighting strategy is adjusted after the site terrain information is detected. The final effect of this type of technology depends on the accuracy of terrain information detection and the granularity of scene classification.

[0004] Currently, scene perception based on GNSS signals mostly uses empirical threshold methods or traditional machine learning algorithms to achieve rough perception of indoor, open outdoor, and outdoor obstructed scenes. The scene granularity is low and it is difficult to meet the complex and changeable scene information requirements in GNSS dynamic positioning.

[0005] Therefore, the fixed GNSS positioning strategy is obviously difficult to adapt to complex and changing dynamic application scenarios. Summary of the invention

[0006] In order to improve the availability, accuracy and continuity of GNSS positioning services in complex scenarios and meet the GNSS positioning needs in dynamic and complex environments, this application provides a GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning.

[0007] The above-mentioned invention objective of the present application is achieved through the following technical solutions:

[0008] A GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning, comprising:

[0009] Acquire GNSS observation data, extract and calculate GNSS data features, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites, and multipath under different preset classification scenarios;

[0010] Constructing a time series GNSS feature vector of the GNSS data feature by using the SFFS algorithm;

[0011] Constructing an LSTM neural network with a time memory function based on the GNSS feature vector;

[0012] Training the LSTM neural network in combination with the GNSS observation data and a scene label preset for a navigation scene to obtain an LSTM model for scene perception;

[0013] Obtain the true position value provided by the navigation device synchronized with the test device, substitute it into the preset inter-satellite-epoch double difference observation equation of pseudorange and carrier phase, and inversely calculate the residual of pseudorange and carrier phase;

[0014] Fitting the functional relationship between the residual and the preset influencing factors to construct the GNSS random models of the different classification scenarios, wherein the preset influencing factors include carrier-to-noise ratio and satellite altitude angle;

[0015] Based on different classification scenarios and GNSS observation data, construct threshold models for the different classification scenarios, wherein parameters of the threshold models include carrier-to-noise ratio, satellite altitude angle, and ratio value;

[0016] After the LSTM model for scene perception receives the current environmental information in real time and determines the classification scene to which the current environment belongs, it calls the GNSS random model under the corresponding classification scene for weight determination, and adjusts the carrier-to-noise ratio, satellite altitude angle and ratio value adapted to the corresponding classification scene through the threshold model.

[0017] Optionally, the acquiring of GNSS observation data, extracting and calculating GNSS data features, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites and multipath in a classification scenario, include:

[0018] A data collection platform is built by using the interest device and the high-precision combined navigation device to collect the number of satellites visible to the GNSS, carrier-to-noise ratio, satellite altitude angle and multipath of the classification scene;

[0019] The multipath acquisition method includes obtaining the difference between the pseudorange change and the Doppler observation value to evaluate the impact of NLOS and multipath in different environments. The specific calculation formula is as follows:

[0020] ; (1)

[0021] In formula (1), is the multipath error; is the difference operator; is the pseudorange observation value; is the Doppler observation value; is the speed of light; is the signal frequency; is the time interval;

[0022] The calculation formula is: , where the subscript , Indicates time.

[0023] Optionally, the acquiring of GNSS observation data, extracting and calculating GNSS data features, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites and multipath in a classification scenario, and further includes:

[0024] The satellite position precision reduction factor PDOP is introduced. When the receiver's spatial rectangular coordinates are , , and clock difference When the positioning calculation is performed as the estimated parameter, the covariance matrix is ​​obtained as follows:

[0025] ; (2)

[0026] The satellite's position precision reduction factor PDOP can be expressed as:

[0027] .

[0028] Optionally, constructing the time series GNSS feature vector of the GNSS data feature by using the SFFS algorithm includes:

[0029] Calculate the carrier-to-noise ratio, satellite altitude angle, and the number of satellites visible to the GNSS at different satellite azimuths and multipath statistics under different preset classification scenarios through a preset statistical algorithm;

[0030] Calculating carrier-to-noise ratio statistics at different satellite altitude angles and different satellite azimuth angles;

[0031] Calculate the satellite altitude angle statistics and the satellite position dilution of precision factor PDOP value under different satellite azimuth angles;

[0032] Constructing a feature set for scene perception based on the number of satellites visible to the GNSS, the multipath statistics, the carrier-to-noise ratio statistics, the satellite altitude angle statistics, and the position dilution of precision factor PDOP value of the satellite;

[0033] The SFFS algorithm is initialized with an empty feature set, and the GNSS data features are added to determine whether the classification performance is improved, and the number of the GNSS data features added each time includes two or more;

[0034] When the classification performance is no longer improved, outputting the current feature set to be used for the GNSS feature vector for navigation scene perception;

[0035] After the GNSS data features are added, if the feature set does not meet the conditions for forming a GNSS feature vector, the GNSS data feature with the worst perception result among the added GNSS data features is deleted. If the deleted GNSS data feature is not the latest one added, the GNSS data feature that does not meet the GNSS feature vector conditions and has the worst perception result will continue to be deleted, otherwise new GNSS data features will continue to be selected for addition.

[0036] Optionally, the LSTM model includes a forget gate, an update gate, and an output gate, the forget gate includes a sigmoid layer, the update gate and the output gate both include a sigmoid layer and a tanh layer, and the LSTM neural network is trained in combination with the GNSS observation data and the scene label preset in the navigation scene to obtain the LSTM model for scene perception, including:

[0037] The calculation formulas of the sigmoid layer and the tanh layer are:

[0038] as well as ;

[0039] In the LSTM model, the expressions of the forget gate, update gate and output gate are:

[0040] Forget Gate:

[0041] ;

[0042] Update Gate:

[0043] ;

[0044] ;

[0045] ;

[0046] Output Gate:

[0047] ;

[0048] ;

[0049] The input information of an LSTM unit in the LSTM model includes the memory cell state of the previous epoch , hidden layer output Characteristic information related to the current epoch , after obtaining the input information, the hidden layer output of the previous epoch is and the current epoch feature information First, the forget gate is entered, and the sigmoid layer of the forget gate determines the forgetting of non-important information, and its output is ;

[0050] and The sigmoid layer of the update gate determines the information that needs to be updated , create new candidate state quantities through the tanh layer , and Multiplication determines the information that needs to be stored in the current cell state, and and Multiplication realizes the discarding of information of the previous epoch;

[0051] and Add together to get the complete cell state of the current epoch , the cell state will be output to the LSTM unit of the next epoch, i.e., the update gate constituting the LSTM model;

[0052] and The sigmoid layer of the output gate determines the output information in the cell state , and the tanh layer passing through the output gate The final output is obtained by multiplication, which constitutes the output gate of the LSTM model.

[0053] Optionally, the step of obtaining the true position value provided by a navigation device synchronized with the test device, substituting the value into a preset inter-satellite-epoch double difference observation equation of pseudorange and carrier phase, and back-calculating the residual of pseudorange and carrier phase includes:

[0054] The GNSS pseudorange observation equation and carrier phase observation equation are as follows:

[0055] ; (3)

[0056] ; (4)

[0057] Among them, the subscript Indicates receiver, superscript Indicates satellite; , are pseudorange observation value and carrier phase observation value respectively; is the geometric distance between the receiver and the satellite, and its calculation formula is: , , are the positions of the receiver and satellite in the ECEF framework, respectively;

[0058] is the speed of light, , are the receiver clock error and the satellite clock error, is the ionospheric delay; is the tropospheric delay; is the multipath error; , , , are respectively a receiver pseudorange, a satellite pseudorange, a receiver phase and a satellite phase hardware delay of the receiver; is the carrier wavelength; is the undifferenced integer ambiguity; , are pseudorange observation noise and carrier phase observation noise respectively;

[0059] Obtain the GNSS observation data of equations (3) and (4) and perform double difference processing between satellites and epochs to construct double difference observation equations;

[0060] First use the reference satellite With the target satellite The inter-satellite single-difference observation equation is constructed by using the observed quantities: ; (5) ; (6)

[0061] In the formula, is the intersatellite single difference operator, , are the inter-satellite differences of pseudorange and carrier phase observations respectively; For receiver and reference satellite , target satellite The difference in geometric distance between Reference satellite Clock error, target satellite The difference in clock error, Reference satellite With the target satellite The difference in ionospheric delay is Reference satellite With the target satellite The difference in tropospheric delay is Reference satellite With the target satellite The difference in multipath error is , Baseline Satellite Pseudorange, target satellite Phase hardware delay intersatellite difference; Reference satellite With the target satellite The inter-satellite single-difference integer ambiguity between ;

[0062] Based on the inter-satellite single-difference observation equation, inter-epoch difference is performed to form the inter-satellite-epoch double-difference observation equation as follows:

[0063] ; (7)

[0064] ; (8)

[0065] In the formula, is the inter-satellite-epoch double difference operator; , It is a time mark; and are the inter-satellite-inter-epoch double difference observations of pseudorange and carrier phase, respectively; is the reference satellite , target satellite At the moment ,time The intersatellite-epoch double difference of the geometric distance between them; is the inter-satellite-epoch double difference of ionospheric delay; is the inter-satellite-epoch double difference of the tropospheric delay; is the inter-satellite-inter-epoch double difference of the multipath error; is the inter-satellite-epoch double-difference integer ambiguity;

[0066] Obtain broadcast ephemeris, UNB3m model and Klobuchar model to calculate satellite clock error ; Get tropospheric delay and ionospheric delay The correction accuracy of the Klobuchar model is 2~8TECU, the satellite clock accuracy is 2.5ns, and the prediction accuracy of the UNB3m model in the zenith direction is 5.2cm;

[0067] When cycle slips do not occur between epochs, after being corrected by the broadcast ephemeris, the UNB3m model and the Klobuchar model, equations (7) and (8) can be simplified to:

[0068] ; (9)

[0069] ; (10)

[0070] Substituting the true value of the receiver coordinates into equations (9) and (10), the a priori residual term containing the multipath error affected by the observation noise and the reaction environment is calculated as follows:

[0071] ; (11)

[0072] ; (12)

[0073] When it is single-frequency data, the single-frequency data includes three systems: GPS, BDS and GALILEO. The double difference residual of single-frequency pseudorange and carrier phase can be expressed as:

[0074] ; (13)

[0075] ; (14)

[0076] In formula (13) and formula (14), , as well as All are known items, and the multipath errors of the three systems of GPS, BDS and GALILEO can be calculated by equations (13) and (14): Pseudorange Observation Noise The comprehensive error term and multipath error Carrier phase observation noise The comprehensive error term According to the error propagation law, the relationship between the mean error of the combined residual and the mean error of the pseudorange and carrier phase undifference residual is:

[0077] ; (15)

[0078] ; (16)

[0079] in, and , and , and , and is the comprehensive error term of adjacent epochs of the same satellite, and equations (15) and (16) can be expressed as:

[0080] ; (17)

[0081] . (18).

[0082] Optionally, fitting the functional relationship between the residual and a preset influencing factor to construct a GNSS random model for different classification scenarios, wherein the preset influencing factor includes the carrier-to-noise ratio and the satellite altitude angle, comprises:

[0083] Select the target satellite The satellite with similar impact factor is the reference satellite For reference, equations (17) and (18) can be expressed as:

[0084] ; (19)

[0085] ; (20)

[0086] Obtain the GNSS observation data of different classification scenarios to calculate , carrier-to-noise ratio, and satellite elevation angle reconstruction of the classified scene pseudo-range random model in different scenarios: ; Among them, CNR is the carrier-to-noise ratio, ELE is the satellite altitude angle, and positioning is based on Calculate the weights of the carrier phase observations.

[0087] Optionally, based on the preset different classification scenarios and GNSS observation data, threshold models for different classification scenarios are constructed, where the threshold models include carrier-to-noise ratio, satellite altitude angle, and ratio value, including:

[0088] Based on the GNSS observation data, the distribution of carrier-to-noise ratio and satellite altitude angle under different classification scenarios is extracted, and the carrier-to-noise ratio and altitude angle thresholds under different classification scenarios are selected according to the triple sigma principle;

[0089] Based on equation (6), the inter-station difference calculation is performed to form the carrier phase station-satellite double difference observation:

[0090] (twenty one)

[0091] In formula (21), the subscript For mobile stations, is the base station; is the carrier phase station-satellite double difference observation value; is the station-satellite double difference of ionospheric delay; is the station-satellite double difference of tropospheric delay; is the station-satellite double difference of multipath error; is the station-satellite double difference integer ambiguity. After substituting into the true coordinates of the receiver, the real station-satellite double difference ambiguity is calculated back as the reference value.

[0092] Perform GNSS RTK positioning, compare the station-satellite double difference integer ambiguity solved by positioning with the real station-satellite double difference ambiguity, determine whether true fixation is achieved, obtain the distribution of the ratio value under different classification scenarios, and output the fixation rate and the correct fixation rate. The correct fixation rate is the fixation rate of the true fixation, and the fixation rate is the fixation rate of the successful fixation. Set the ratio threshold according to the fixation rate and the correct fixation rate.

[0093] The above-mentioned second objective of the present application is achieved through the following technical solutions:

[0094] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the above-mentioned GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning are implemented.

[0095] The third objective of the present application is achieved through the following technical solutions:

[0096] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning.

[0097] In summary, the present application includes at least one of the following beneficial technical effects:

[0098] 1. The present invention proposes an adaptive GNSS positioning model with enhanced navigation scene perception. When only a single GNSS signal source is used, the time-related characteristics of GNSS feature information are considered to construct a multi-dimensional GNSS feature vector. Based on a large number of measured GNSS observation data of complex scenes, a long short-term memory LSTM neural network is designed and trained to realize outdoor complex navigation scene perception. Based on the set classification scene and a large number of GNSS observation data, the carrier-to-noise ratio, altitude angle, and ratio threshold models of the classification scene are constructed, and the coordinate true value is used to bring into the pseudorange / carrier phase double difference observation equation to inversely calculate the GNSS observation value residual, which is used to construct the classification scene GNSS random model. Combined with the environmental information provided by the LSTM neural network and the classification scene threshold model, the threshold and weight information provided by the GNSS random model, the adaptive optimization of the GNSS model with enhanced navigation scene perception is realized to meet the GNSS positioning requirements in dynamic and complex environments.

[0099] 2. Considering that there is a certain connectivity between navigation scene epochs, that is, the current epoch scene is more likely to be consistent with the previous epoch scene, the recurrent neural network (RNN) with current features and previous features as common input is more suitable for navigation scene perception. Since RNN introduces the concept of hidden layer, the neurons with self-feedback ability contained in it can be used to process time series signals of any length. As a variant of RNN, LSTM further adds a gating structure, which better solves the problem of gradient disappearance that may occur in long-term perception, and is better than RNN. Therefore, LSTM is selected for navigation scene perception;

[0100] 3. In complex environments, multipath is serious, and the pseudo-range measurement error caused by it may reach several meters or even more than ten meters, while the maximum impact of multipath error on carrier phase is only at the centimeter level. By comparison, the main influencing term in formula (9) is Therefore, the pseudorange observation residual can be calculated by formula (9) to construct the random model of pseudorange classification scenario; while formula (10) may also include the influence of other error terms, but since the observation environment of the carrier and pseudorange is the same, after reconstructing the random model of pseudorange classification scenario and reflecting the influence of the environment on the quality of GNSS observations, the residual ratio of pseudorange and carrier phase observations can be calculated, and the corresponding ratio can be used to obtain the specific weight of the carrier phase observation in the corresponding scenario and GNSS positioning.

[0101] 4. When constructing a random model, the carrier-to-noise ratio or satellite altitude angle is usually selected as the influencing factor. Among them, the height of the satellite altitude angle mainly reflects the strength of atmospheric errors such as tropospheric delay, and the size of the carrier-to-noise ratio mainly reflects the diffraction error caused by the observation environment during the satellite signal transmission process. Both are commonly used random model influencing factors; the technical solution fully utilizes the combined advantages of the two, selects two influencing factors to jointly construct a random model, and after selecting the influencing factors, it is believed that the accuracy of GNSS observations obeys a certain function distribution of the influencing factors, that is, the observations with the same influencing factors have the same accuracy. Therefore, the target satellite is selected. Satellites with similar impact factors as possible are used as reference satellites. . BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 This is a flow chart of the SFFS algorithm in an embodiment of a GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning of the present application;

[0103] Figure 2 This is a flowchart of an implementation of an embodiment of a GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning of the present application;

[0104] Figure 3 It is a structural schematic diagram of a computer device of the present application. DETAILED DESCRIPTION

[0105] The present application is further described in detail below in conjunction with the accompanying drawings.

[0106] In one embodiment, if Figure 1-3 As shown, the present application discloses a GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning, which specifically includes the following steps:

[0107] S1: Acquire GNSS observation data, extract and calculate GNSS data features, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites and multipath under different preset classification scenarios;

[0108] In this embodiment, a platform for collecting GNSS observation data is composed of interest devices and high-precision combined navigation devices. The interest devices include Jinwei DM711 devices, which provide a data basis for the construction of scene perception models and GNSS positioning models.

[0109] GNSS signal characteristics can be divided into four categories: satellite visibility, satellite signal strength, multipath, and satellite geometric distribution. Based on these four categories, GNSS feature vectors with differences between scenes are constructed. Specifically, the number of GNSS visible satellites, signal-to-noise ratio, and satellite altitude angles in different classification scenes are collected. In addition, considering that the signal obstruction in the urban canyon scene mainly comes from both sides of the road, and the signal obstruction in the elevated scene in different classification scenes mainly comes from the zenith direction, etc., that is, the satellite azimuth also contains scene information. Therefore, the satellite azimuth is calculated in combination with the carrier attitude information provided by the combined navigation device. In addition, the difference between the pseudorange change and the Doppler observation value is used to evaluate the impact of NLOS and multipath in different environments. The specific calculation formula for obtaining the multipath is as follows:

[0110] ; (1)

[0111] In formula (1), is the multipath error; is the difference operator; is the pseudorange observation value; is the Doppler observation value; is the speed of light; is the signal frequency; For the time interval. The calculation formula is: , where the subscript , Indicates time.

[0112] In order to evaluate the satellite geometric distribution, the satellite position precision dilution factor PDOP is introduced to reflect the satellite geometric strength while counting the various satellite elevation angle statistics. , , and clock difference When the positioning calculation is performed as the estimated parameter, the covariance matrix is ​​obtained as follows:

[0113] ; (2)

[0114] The satellite position precision reduction factor PDOP can be expressed as: .

[0115] S2: constructing a time series GNSS feature vector of the GNSS data feature by using the SFFS algorithm;

[0116] In this embodiment, the carrier-to-noise ratio, satellite altitude angle, and the number of satellites visible to the GNSS and multipath statistics under different preset classification scenarios are calculated by combining statistical methods such as range, quartile, interquartile range, mean, standard deviation, and ratio (current value / scenario maximum value);

[0117] Calculating carrier-to-noise ratio statistics at different satellite altitude angles and different satellite azimuth angles;

[0118] Calculate the satellite altitude angle statistics and the satellite position dilution of precision factor PDOP value under different satellite azimuth angles;

[0119] A feature set for scene perception is constructed based on the number of satellites visible to the GNSS, the multipath statistics, the carrier-to-noise ratio statistics, the satellite altitude angle statistics, and the position dilution of precision factor PDOP value of the satellite.

[0120] The above calculated values ​​are used to construct a feature set for scene perception, and the SFFS (Sequential Forward Floating Selection) algorithm is used to select the feature values ​​and construct the final feature vector. The SFFS algorithm process is as follows: Figure 2 shown.

[0121] The SFFS algorithm is initialized with an empty feature set, and GNSS data features are added to determine whether the classification performance is improved. The number of GNSS data features added each time includes two or more;

[0122] When the classification performance is no longer improved, outputting the current feature set to be used for the GNSS feature vector for navigation scene perception;

[0123] After adding GNSS data features, if the feature set does not meet the conditions for forming a GNSS feature vector, delete the GNSS data feature with the worst perception result among the added GNSS data features. If the deleted GNSS data feature is not the latest one added, continue to delete the GNSS data feature that does not meet the GNSS feature vector conditions and has the worst perception result. Otherwise, continue to select new GNSS data features to add.

[0124] S3: constructing an LSTM neural network with time memory function based on the GNSS feature vector;

[0125] S4: Train the LSTM neural network by combining GNSS observation data and scene labels preset in navigation scenarios to obtain an LSTM model for scene perception;

[0126] In this embodiment, the LSTM unit is shown in Table 1 below. The LSTM model includes a forget gate, an update gate, and an output gate. The forget gate includes a sigmoid layer. The update gate and the output gate both include a sigmoid layer and a tanh layer. In Table 1 below, the calculation formulas for the sigmoid layer and the tanh layer are: , .

[0127] Table 1: LSTM unit

[0128]

[0129] Specifically, the input of an LSTM unit in the LSTM model includes the memory cell state of the previous epoch , hidden layer output Characteristic information related to the current epoch After obtaining the above input, the hidden layer output of the previous epoch is and the current epoch feature information First, it enters the forget gate, and the sigmoid layer of the forget gate determines the forgetting of non-important information. Its output is ;

[0130] Then, and The sigmoid layer of the update gate determines the information that needs to be updated , create new candidate state quantities through the tanh layer , and Multiplication determines the information that needs to be stored in the current cell state, and and Multiplication realizes the discarding of information of the previous epoch,

[0131] and Add together to get the complete cell state of the current epoch , the cell state will be output to the LSTM unit of the next epoch, which constitutes the update gate of the LSTM;

[0132] and The sigmoid layer of the output gate determines the output information in the cell state , And the tanh layer after the output gate The final output is obtained by multiplication, which constitutes the output gate of LSTM. LSTM solves the problem of gradient disappearance in the long-term memory process of neural network through the above gate control structure.

[0133] In order to obtain accurate and reliable navigation scene perception results, set LSTM hyperparameters suitable for GNSS feature vectors. There are three main LSTM hyperparameters to be set: the number of hidden layers, the number of hidden layer neurons, and the time window. Generally speaking, the more hidden layers, the more hidden layer neurons, and the longer the time window, the stronger the model's expressive ability, but also means longer model training time and a greater risk of model overfitting. Therefore, under the premise of taking into account the training accuracy / loss, test accuracy / loss, and training time, the hyperparameters are set as follows: 2 hidden layers, 26 hidden layer neurons, and 7s time window.

[0134] S5: Obtain the true position value provided by the navigation device synchronized with the test device, substitute it into the preset inter-satellite-epoch double difference observation equation of pseudorange and carrier phase, and inversely calculate the residual of pseudorange and carrier phase;

[0135] The GNSS pseudorange observation equation and carrier phase observation equation are as follows:

[0136] ; (3)

[0137] ; (4)

[0138] Among them, the subscript Indicates the receiver ( Can be a mobile station or base station ); superscript Indicates satellite; , are pseudorange observation value and carrier phase observation value respectively; is the geometric distance between the receiver and the satellite, and its calculation formula is: , , are the positions of the receiver and satellite in the ECEF framework, respectively; is the speed of light; , are the receiver clock error and the satellite clock error respectively; is the ionospheric delay; is the tropospheric delay; is the multipath error; , , , They are the receiver pseudorange, satellite pseudorange, receiver phase and satellite phase hardware delay of the receiver respectively; is the carrier wavelength; is the undifferenced integer ambiguity; , They are pseudorange observation noise and carrier phase observation noise respectively.

[0139] In order to obtain relatively pure observations that only contain environmental information (multipath error) and observation noise, the GNSS random model for classification scenarios is constructed using them. The double-difference observation equation is constructed by using the original observation values ​​of equations (3) and (4) and the GNSS observation data of equations (3) and (4) to perform inter-satellite and inter-epoch double-difference processing.

[0140] Using reference satellites With the target satellite The inter-satellite single-difference observation equation is constructed by using the observed quantities:

[0141] ; (5)

[0142] ; (6)

[0143] In the formula, is the intersatellite single difference operator, , are the inter-satellite differences of pseudorange and carrier phase observations respectively; For receiver and reference satellite , target satellite The difference in geometric distance between Reference satellite Clock error, target satellite The difference in clock error, Reference satellite With the target satellite The difference in ionospheric delay is Reference satellite With the target satellite The difference in tropospheric delay is Reference satellite With the target satellite The difference in multipath error is , Baseline Satellite Pseudorange, target satellite Phase hardware delay intersatellite difference; Reference satellite With the target satellite The inter-satellite single-difference integer ambiguity between them.

[0144] In order to further weaken the spatial correlation errors such as satellite-side errors, atmospheric errors (ionospheric delay, tropospheric delay), orbit errors, etc., the inter-epoch difference is performed on the basis of the inter-satellite difference to form the inter-satellite-epoch double difference observation equation as follows:

[0145] ; (7)

[0146] ; (8)

[0147] In the formula, is the inter-satellite-epoch double difference operator; , It is a time mark; and are the inter-satellite-inter-epoch double difference observations of pseudorange and carrier phase, respectively; is the reference satellite , target satellite At the moment ,time The intersatellite-epoch double difference of the geometric distance between them; is the inter-satellite-epoch double difference of ionospheric delay; is the inter-satellite-epoch double difference of the tropospheric delay; is the inter-satellite-inter-epoch double difference of the multipath error; is the inter-satellite-epoch double-difference integer ambiguity.

[0148] Calculate satellite clock errors using broadcast ephemeris, UNB3m model based on weather forecasts, and Klobuchar model , Tropospheric delay Ionospheric delay According to the product information released by IGS, the correction accuracy of the Klobuchar model is 2~8TECU, the satellite clock accuracy is 2.5ns, and the prediction accuracy of the UNB3m model in the zenith direction is 5.2cm. , ionospheric delay , Tropospheric delay , Satellite hardware delay All are slow variables, and no cycle slip occurs between epochs. After being corrected by the broadcast ephemeris, UNB3m model and Klobuchar model, equations (7) and (8) can be simplified to:

[0149] ; (9)

[0150] ; (10)

[0151] It is worth noting that in complex environments, multipath is serious, and the pseudo-range measurement error caused by it may reach several meters or even more than ten meters, while the maximum impact of multipath error on carrier phase is only at the centimeter level. By comparison, the main influencing term in formula (9) is Therefore, the pseudorange observation residual can be calculated by formula (9) to construct the random model of pseudorange classification scenario; while formula (10) may also include the influence of other error terms, but since the observation environment of the carrier and pseudorange is the same, after reconstructing the random model of pseudorange classification scenario and reflecting the influence of the environment on the quality of GNSS observations, the residual ratio of pseudorange and carrier phase observations can be calculated, and the corresponding ratio can be used to obtain the specific weight of the carrier phase observation in the corresponding scenario and GNSS positioning.

[0152] Substituting the true value of the receiver coordinates into equations (9) and (10), the pre-test residual term, which mainly includes the multipath error of the observation value noise and the reaction environment, is calculated as follows:

[0153] (11)

[0154] (12)

[0155] In this embodiment, taking GPS, BDS and GALILEO single-frequency data as an example, the double difference residuals of the single-frequency pseudorange and carrier phase of the three systems can be expressed as:

[0156] ; (13)

[0157] ; (14)

[0158] Since in formula (13) and formula (14), , , All are known items, and the multipath errors of the three systems can be calculated by equations (13) and (14): Pseudorange Observation Noise The comprehensive error term , multipath error Carrier phase observation noise The comprehensive error term According to the error propagation law, the relationship between the mean error of the above combined residual and the mean error of the pseudorange and carrier phase undifference residual is:

[0159] ; (15)

[0160] ; (16)

[0161] in, and , and , and , and is the comprehensive error term of adjacent epochs of the same satellite. Since the time interval between adjacent epochs of GNSS is generally less than or equal to 1 s, it is assumed that the comprehensive error change of the same satellite within 1 s is very small. Therefore, equations (15) and (16) can be expressed as:

[0162] ; (17)

[0163] . (18)

[0164] S6: Fitting the functional relationship between the residual and the preset influencing factors to construct the GNSS random models of the different classification scenarios, wherein the preset influencing factors include carrier-to-noise ratio and satellite altitude angle;

[0165] In this embodiment, the carrier-to-noise ratio or the satellite altitude angle is usually selected as the influencing factor when constructing the random model. Among them, the height of the satellite altitude angle mainly reflects the strength of atmospheric errors such as tropospheric delay, and the size of the carrier-to-noise ratio mainly reflects the diffraction error caused by the observation environment during the satellite signal transmission process. Both are commonly used random model influencing factors. The present invention makes full use of the combined advantages of the two and selects two influencing factors to jointly construct a random model. After selecting the influencing factors, it is believed that the accuracy of GNSS observation values ​​obeys a certain function distribution of the influencing factors, that is, the observation values ​​with the same influencing factors have the same accuracy. Therefore, the selected influencing factors are related to the target satellite. The satellite with the closest influence factor as possible is used as the reference satellite, that is, as the benchmark satellite , at this time, equation (17) and equation (18) can be expressed as:

[0166] ; (19)

[0167] ; (20)

[0168] Calculate using a large number of GNSS observation data of different classification scenarios , Carrier-to-Noise Ratio , Satellite altitude angle Reconstruct the classification scene pseudo-range random model under different scenarios: ; Among them, CNR is the carrier-to-noise ratio, ELE is the satellite altitude angle, and positioning is based on Calculate the weights of carrier phase observations

[0169] S7: constructing threshold models of different classification scenarios based on different classification scenarios and GNSS observation data, wherein parameters of the threshold models include carrier-to-noise ratio, satellite altitude angle, and ratio value;

[0170] In this embodiment, based on a large amount of GNSS observation data, the distribution of the carrier-to-noise ratio CNR and the satellite elevation angle ELE in different classification scenarios are extracted, and the carrier-to-noise ratio CNR and the elevation angle ELE thresholds in different classification scenarios are selected according to the triple sigma principle.

[0171] In order to obtain the appropriate ratio threshold in different classification scenarios, the inter-station difference calculation is performed based on formula (6) to form the carrier phase station-satellite double difference observation:

[0172] (twenty one)

[0173] Through the inter-station difference, satellite-side correlation error, satellite clock error Delay with satellite hardware is completely eliminated. In formula (21), the subscript For mobile stations, is the base station; is the carrier phase station-satellite double difference observation value; is the station-satellite double difference of ionospheric delay; is the station-satellite double difference of tropospheric delay; is the station-satellite double difference of multipath error; is the station-satellite double difference integer ambiguity. After substituting into the true coordinates of the receiver, the real station-satellite double difference ambiguity is calculated back as the reference value.

[0174] Perform GNSS RTK positioning, compare the station-satellite double difference integer ambiguity solved by positioning with the real station-satellite double difference ambiguity, determine whether true fixation is achieved, obtain the distribution of the ratio value under different classification scenarios and output the fixation rate and correct fixation rate. The correct fixation rate is the fixation rate of true fixation, and the fixation rate is the fixation rate of successful fixation. Set the ratio threshold according to the fixation rate and the correct fixation rate.

[0175] S8: After the LSTM model used for scene perception receives the current environmental information in real time and determines the classification scene to which the current environment belongs, it calls the GNSS random model under the corresponding classification scene for weight determination, and adjusts the carrier-to-noise ratio, satellite altitude angle, and ratio value adapted to the corresponding classification scene through the threshold model.

[0176] In one embodiment, referring to Figure 2, extract and calculate the GNSS data features such as carrier-to-noise ratio, satellite altitude angle, DOP, number of satellites, etc. in the scene of interest through a large amount of GNSS observation data. According to the connectivity between navigation scenes, the time series GNSS feature vector is constructed through the SFFS algorithm, and the hyperparameters of the LSTM model with time memory are designed accordingly. The scene perception model is obtained by training the LSTM with a large amount of measured GNSS data combined with the set scene labels. The true position value provided by the high-precision combined navigation device synchronously mounted with the test equipment is obtained, and the pseudorange and carrier phase are inverted into the inter-satellite-epoch double difference observation equation. The pseudorange and carrier phase residuals are inverted, and the two influencing factors of carrier-to-noise ratio and satellite altitude angle are comprehensively considered. The functional relationship between the residual and the influencing factor is fitted to obtain the GNSS random model of the classification scene; the CNR and ELE thresholds in different scenes are set according to the triple sigma principle; the true position value is brought into the station-satellite double difference observation equation, and the real double difference ambiguity is inverted. The ratio thresholds of different scenes are set by comprehensively considering the fixation rate and the correct fixation rate. After obtaining the LSTM model and the GNSS positioning model, the LSTM model is used to obtain environmental information during the positioning process, and the random model and the threshold model are adjusted accordingly to achieve adaptive optimization of the GNSS positioning model with enhanced navigation scene perception.

[0177] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0178] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning is implemented.

[0179] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning is implemented.

[0180] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, a GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning is implemented.

[0181] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0182] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0183] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A GNSS model adaptive optimization method for enhanced navigation scene perception in high-precision positioning, characterized in that: include: Acquire GNSS observation data, extract and calculate GNSS data features, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites, and multipath under different preset classification scenarios; Constructing a time series GNSS feature vector of the GNSS data feature by using the SFFS algorithm; Constructing an LSTM neural network with a time memory function based on the GNSS feature vector; Training the LSTM neural network in combination with the GNSS observation data and the scene labels preset for the navigation scene to obtain an LSTM model for scene perception; Obtain the true position value provided by the navigation device synchronized with the test device, substitute it into the preset inter-satellite-epoch double difference observation equation of pseudorange and carrier phase, and inversely calculate the residual of pseudorange and carrier phase; Fitting the functional relationship between the residual and the preset influencing factors to construct the GNSS random models of the different classification scenarios, wherein the preset influencing factors include carrier-to-noise ratio and satellite altitude angle; Based on different classification scenarios and GNSS observation data, construct threshold models for the different classification scenarios, wherein parameters of the threshold models include carrier-to-noise ratio, satellite altitude angle, and ratio value; After the LSTM model for scene perception receives the current environment information in real time and determines the classification scene to which the current environment belongs, the GNSS random model under the corresponding classification scene is called for weight determination, and the carrier-to-noise ratio, satellite altitude angle and ratio value adapted to the corresponding classification scene are adjusted through the threshold model; The method of obtaining the true position value provided by the navigation device synchronized with the test device, substituting the value into the preset inter-satellite-epoch double difference observation equation of pseudorange and carrier phase, and back-calculating the residual of pseudorange and carrier phase includes: The GNSS pseudorange observation equation and carrier phase observation equation are as follows: ； （3） ； （4） Among them, the subscript Indicates receiver, superscript Indicates satellite; , are pseudorange observation value and carrier phase observation value respectively; is the geometric distance between the receiver and the satellite, and its calculation formula is: , , are the positions of the receiver and satellite in the ECEF framework, respectively; is the speed of light, , are the receiver clock error and the satellite clock error, is the ionospheric delay; is the tropospheric delay; is the multipath error; , , , are respectively a receiver pseudorange, a satellite pseudorange, a receiver phase and a satellite phase hardware delay of the receiver; is the carrier wavelength; is the undifferenced integer ambiguity; , are pseudorange observation noise and carrier phase observation noise respectively; Obtain the GNSS observation data of equations (3) and (4) and perform double difference processing between satellites and epochs to construct double difference observation equations; First use the reference satellite With the target satellite The inter-satellite single-difference observation equation is constructed by using the observed quantities: ; (5) ; (6) In the formula, is the intersatellite single difference operator, , are the inter-satellite differences of pseudorange and carrier phase observations respectively; For receiver and reference satellite , target satellite The difference in geometric distance between Reference satellite Clock error, target satellite The difference in clock error, Reference satellite With the target satellite The difference in ionospheric delay is Reference satellite With the target satellite The difference in tropospheric delay is Reference satellite With the target satellite The difference in multipath error is , Baseline Satellite With the target satellite Pseudorange and phase hardware delay inter-satellite difference; Reference satellite With the target satellite The inter-satellite single-difference integer ambiguity between ; Based on the inter-satellite single-difference observation equation, inter-epoch difference is performed to form the inter-satellite-epoch double-difference observation equation as follows: ； （7） ；（8） In the formula, is the inter-satellite-epoch double difference operator; , It is a time mark; and are the inter-satellite-inter-epoch double difference observations of pseudorange and carrier phase, respectively; is the reference satellite , target satellite At the moment ,time The intersatellite-epoch double difference of the geometric distance between them; is the inter-satellite-epoch double difference of ionospheric delay; is the inter-satellite-epoch double difference of the tropospheric delay; is the inter-satellite-inter-epoch double difference of the multipath error; is the inter-satellite-epoch double-difference integer ambiguity; Obtain broadcast ephemeris, UNB3m model and Klobuchar model to calculate satellite clock error ; Get tropospheric delay and ionospheric delay The correction accuracy of the Klobuchar model is 2~8TECU, the satellite clock accuracy is 2.5ns, and the prediction accuracy of the UNB3m model in the zenith direction is 5.2cm; When cycle slips do not occur between epochs, after being corrected by the broadcast ephemeris, the UNB3m model and the Klobuchar model, equations (7) and (8) can be simplified to: ； （9） ； （10） Substituting the true value of the receiver coordinates into equations (9) and (10), the a priori residual term containing the multipath error affected by the observation noise and the reaction environment is calculated as follows: ； （11） ； （12） When it is single-frequency data, the single-frequency data includes three systems: GPS, BDS and GALILEO. The double difference residual of single-frequency pseudorange and carrier phase can be expressed as: ； （13） ；（14） In formula (13) and formula (14), , as well as All are known items, and the multipath errors of the three systems of GPS, BDS and GALILEO can be calculated by equations (13) and (14): Pseudorange Observation Noise The comprehensive error term and multipath error Carrier phase observation noise The comprehensive error term According to the error propagation law, the relationship between the mean error of the combined residual and the mean error of the pseudorange and carrier phase undifference residual is: ； （15） ； （16） in, and , and , and , and is the comprehensive error term of adjacent epochs of the same satellite, and equations (15) and (16) can be expressed as: ； （17） ； （18） The fitting of the functional relationship between the residual and the preset influencing factors to construct GNSS random models for different classification scenarios, wherein the preset influencing factors include the carrier-to-noise ratio and the satellite altitude angle, comprises: Select the target satellite The satellite with similar impact factor is the reference satellite For reference, equations (17) and (18) can be expressed as: ； （19） ； （20） Obtain the GNSS observation data of different classification scenarios to calculate , carrier-to-noise ratio, and satellite elevation angle reconstruction of the classified scene pseudo-range random model in different scenarios: ; Among them, CNR is the carrier-to-noise ratio, ELE is the satellite altitude angle, and positioning is based on Calculate the weights of the carrier phase observations.

2. The GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning according to claim 1 is characterized in that: The GNSS observation data is obtained, and GNSS data features are extracted and calculated, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites and multipath in the classification scenario, including: A data collection platform is built by using the interest device and the high-precision combined navigation device to collect the number of satellites visible to the GNSS, carrier-to-noise ratio, satellite altitude angle and multipath of the classification scene; The multipath acquisition method includes obtaining the difference between the pseudorange change and the Doppler observation value to evaluate the impact of NLOS and multipath in different environments. The specific calculation formula is as follows: ； （1） In formula (1), is the multipath error; is the difference operator; is the pseudorange observation value; is the Doppler observation value; is the speed of light; is the signal frequency; is the time interval; The calculation formula is: , where the subscript , Indicates time.

3. The GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning according to claim 2 is characterized in that: The GNSS observation data is obtained, and GNSS data features are extracted and calculated, wherein the GNSS data features include carrier-to-noise ratio, satellite altitude angle, DOP data, number of satellites and multipath in the classification scenario, and also include: The satellite position precision reduction factor PDOP is introduced. When the receiver's spatial rectangular coordinates are , , and clock difference When the positioning calculation is performed as the estimated parameter, the covariance matrix is ​​obtained as follows: ； （2） The satellite's position precision reduction factor PDOP can be expressed as: 。 4. The GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning according to claim 3 is characterized in that: The step of constructing the time series GNSS feature vector of the GNSS data feature by using the SFFS algorithm includes: Calculate the carrier-to-noise ratio, satellite altitude angle, and the number of satellites visible to the GNSS at different satellite azimuths and multipath statistics under different preset classification scenarios through a preset statistical algorithm; Calculating carrier-to-noise ratio statistics at different satellite altitude angles and different satellite azimuth angles; Calculate the satellite altitude angle statistics and the satellite position dilution of precision factor PDOP value under different satellite azimuth angles; Constructing a feature set for scene perception based on the number of satellites visible to the GNSS, the multipath statistics, the carrier-to-noise ratio statistics, the satellite altitude angle statistics, and the position dilution of precision factor PDOP value of the satellite; The SFFS algorithm is initialized with an empty feature set, and the GNSS data features are added to determine whether the classification performance is improved, and the number of the GNSS data features added each time includes two or more; When the classification performance is no longer improved, outputting the current feature set to be used for the GNSS feature vector for navigation scene perception; After the GNSS data features are added, if the feature set does not meet the conditions for forming a GNSS feature vector, the GNSS data feature with the worst perception result among the added GNSS data features is deleted. If the deleted GNSS data feature is not the latest one added, the GNSS data feature that does not meet the GNSS feature vector conditions and has the worst perception result will continue to be deleted, otherwise new GNSS data features will continue to be selected for addition.

5. The GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning according to claim 1 is characterized in that: The LSTM model includes a forget gate, an update gate, and an output gate, the forget gate includes a sigmoid layer, the update gate and the output gate both include a sigmoid layer and a tanh layer, and the LSTM neural network is trained by combining GNSS observation data and scene labels preset in navigation scenes to obtain an LSTM model for scene perception, including: The calculation formulas of the sigmoid layer and the tanh layer are: as well as ; In the LSTM model, the expressions of the forget gate, update gate and output gate are: Forget Gate: ； Update Gate: ； ； ； Output Gate: ； ； The input information of an LSTM unit in the LSTM model includes the memory cell state of the previous epoch , hidden layer output Characteristic information related to the current epoch , after obtaining the input information, the hidden layer output of the previous epoch is and the current epoch feature information First, the forget gate is entered, and the sigmoid layer of the forget gate determines the forgetting of non-important information, and its output is ; and The sigmoid layer of the update gate determines the information that needs to be updated , create new candidate state quantities through the tanh layer , and Multiplication determines the information that needs to be stored in the current cell state, and and Multiplication realizes the discarding of information of the previous epoch; and Add together to get the complete cell state of the current epoch , the cell state will be output to the LSTM unit of the next epoch, i.e., the update gate constituting the LSTM model; and The sigmoid layer of the output gate determines the output information in the cell state , and the tanh layer passing through the output gate The final output is obtained by multiplication, which constitutes the output gate of the LSTM model.

6. The GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning according to claim 1 is characterized in that: Based on the preset different classification scenarios and GNSS observation data, threshold models for different classification scenarios are constructed, and the threshold models include carrier-to-noise ratio, satellite altitude angle and ratio value, including: Based on the GNSS observation data, the distribution of carrier-to-noise ratio and satellite altitude angle in different classification scenarios is extracted, and the carrier-to-noise ratio and altitude angle thresholds in different classification scenarios are selected according to the triple sigma principle; Based on equation (6), the inter-station difference calculation is performed to form the carrier phase station-satellite double difference observation: （21） In formula (21), the subscript For mobile stations, is the base station; is the carrier phase station-satellite double difference observation value; is the station-satellite double difference of ionospheric delay; is the station-satellite double difference of tropospheric delay; is the station-satellite double difference of multipath error; is the station-satellite double difference integer ambiguity. After substituting into the true coordinates of the receiver, the real station-satellite double difference ambiguity is calculated back as the reference value. Perform GNSS RTK positioning, compare the station-satellite double difference integer ambiguity solved by positioning with the real station-satellite double difference ambiguity, determine whether true fixation is achieved, obtain the distribution of the ratio value under different classification scenarios, and output the fixation rate and the correct fixation rate. The correct fixation rate is the fixation rate of the true fixation, and the fixation rate is the fixation rate of the successful fixation. Set the ratio threshold according to the fixation rate and the correct fixation rate.

7. A computer 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 computer program, the steps of the GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning as described in any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of a GNSS model adaptive optimization method for enhancing navigation scene perception in high-precision positioning as described in any one of claims 1 to 6 are implemented.

Images