A method for enhancing GNSS positioning accuracy

Abstract

The present invention discloses a method for enhancing GNSS positioning accuracy, comprising the following steps: step S1, constructing a GNSS positioning function based on a mobile station, a base station and a GNSS satellite system; step S2, mapping and approximating the GNSS positioning function using a neural network, and converting the GNSS positioning function into a GNSS precise positioning model to enhance GNSS positioning accuracy. The present invention utilizes a GNSS combined positioning system that integrates a GPS satellite system, a BDS satellite system, a GLONASS satellite system and a Galileo satellite system, can obtain more available satellites, improve the spatial geometric structure of the satellite, thereby improving the availability, reliability and accuracy of the GNSS satellite system, and performs pseudo-range differential positioning on the GNSS satellite system to obtain a positioning function for the position of the GNSS receiver, and uses a neural network to perform nonlinear mapping on the positioning function to obtain a GNSS precise positioning model that fully matches the mapping relationship between satellite positioning data, and has high positioning operation efficiency and high accuracy.

Description

Technical Field

[0001] The present invention relates to the field of satellite positioning technology, and in particular to a method for enhancing GNSS positioning accuracy. Background Art

[0002] A Global Navigation Satellite System (GNSS) generally refers to a system that allows a position fix to be determined based on GNSS signals received from multiple GNSS satellites. Each GNSS satellite transmits a GNSS signal that identifies the satellite and time of signal transmission. The GNSS antenna / receiver is configured to receive each of the GNSS signals transmitted by visible GNSS satellites, and uses the flight time of each GNSS signal and the known position of each GNSS satellite to determine the pseudorange or distance from the GNSS antenna / receiver to the corresponding GNSS satellite. The multiple calculated pseudoranges are used to trilaterate the position of the GNSS antenna / receiver in three-dimensional space. Types of GNSS systems include Global Positioning System (GPS), GLONASS, Galileo, BeiDou (BDS), etc.

[0003] Prior art CN202011024660.5 discloses a system and method for position estimation using two or more antennas, including a first GNSS antenna / receiver, a second GNSS antenna / receiver and a GNSS processor system. The first GNSS antenna / receiver is located at a first position and calculates a first pseudorange based on the received GNSS signal. The second GNSS antenna / receiver is located at a second position at a known distance from the first GNSS antenna / receiver, wherein the second GNSS antenna / receiver calculates a second pseudorange based on the received GNSS signal. The GNSS processor system is configured to receive the first pseudorange and the second pseudorange, wherein in response to the GNSS processor system identifying one of the first and second pseudoranges as erroneous and identifying one of the first and second pseudoranges as valid, the GNSS processing system calculates a corrected pseudorange and uses the corrected pseudorange and the valid pseudorange to determine the GNSS position positioning estimate of the first GNSS antenna / receiver and the second GNSS antenna / receiver, but the positioning estimate is performed by only relying on the pseudorange data to screen and correct in sequence, the data has high viscosity, once one data is wrong, the positioning estimate will fail, and it is difficult to completely match the mapping relationship between satellite positioning data using a fixed mathematical function for positioning solution, and the overall accuracy and efficiency are low. Summary of the invention

[0004] The purpose of the present invention is to provide a GNSS positioning accuracy enhancement method to solve the technical problems in the prior art that positioning estimation is performed by only relying on pseudo-range data to be screened and corrected in sequence, the data has high data adhesion, once one data is wrong, the positioning estimation will fail, and it is difficult to completely match the mapping relationship between satellite positioning data using a fixed mathematical function for positioning solution, resulting in low overall accuracy and low efficiency.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0006] A method for enhancing GNSS positioning accuracy comprises the following steps:

[0007] Step S1, constructing a GNSS positioning function based on the mobile station, the base station and the GNSS satellite system;

[0008] Step S2: Mapping and approximating the GNSS positioning function using a neural network, and converting the GNSS positioning function into a GNSS precise positioning model to enhance the GNSS positioning accuracy.

[0009] As a preferred solution of the present invention, in step S1, the method for constructing the GNSS positioning function includes:

[0010] Step S101: construct a reference pseudorange observation equation based on a reference station and a GNSS satellite system, wherein the reference pseudorange observation equation is characterized by pseudorange observation values ​​from the reference station to all satellites in the GNSS satellite system, wherein the reference pseudorange observation equation is:

[0011]

[0012] Among them, B represents the base station, j represents the satellite, is the pseudorange observation value from the reference station to satellite j, is the geometric distance from the reference station to satellite j, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the reference station, is the pseudorange observation noise and other non-modeled errors, (X B ,Y B ,Z B ) is the three-dimensional coordinate of the base station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite at the moment when satellite j transmits signal data;

[0013] Step S102: construct a mobile pseudorange observation equation based on the mobile station and the GNSS satellite system, wherein the reference pseudorange observation equation is characterized by pseudorange observation values ​​from the mobile station to all satellites in the GNSS satellite system, wherein the mobile pseudorange observation equation is:

[0014] f2=P M j =ρ M j +δt M j +I M j +T M j +ε M j ;

[0015] Among them, M represents the mobile station, j represents the satellite, is the pseudorange observation value from the mobile station to satellite j, is the geometric distance from the mobile station to satellite j, δt M j is the clock difference matrix parameter between the mobile station receiver and the satellite, are the ionospheric error and tropospheric error of the mobile station, is the pseudorange observation noise and other non-modeled errors;

[0016] Step S103: constructing a GNSS positioning function based on the reference pseudorange observation equation and the mobile pseudorange observation equation, wherein the GNSS positioning function is characterized by a nonlinear mapping relationship between the three-dimensional coordinates of the mobile station and the pseudorange observation values ​​and the three-dimensional coordinates of all satellites in the GNSS satellite system.

[0017] As a preferred solution of the present invention, step S101 further includes correcting the reference pseudorange observation equation to obtain a pseudorange correction equation, wherein the pseudorange correction equation is characterized by a pseudorange correction value from the reference station to all satellites in the GNSS satellite system, wherein the pseudorange correction equation is:

[0018] Δρ B j =P B j -ρ B j =δt B j +I B j +T B j +ε B j ;

[0019] Among them, B represents the base station, j represents the satellite, is the pseudorange observation value from the reference station to satellite j, is the geometric distance from the reference station to satellite j, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the reference station, is the pseudorange observation noise and other non-modeled errors, (X B ,Y B ,Z B ) is the three-dimensional coordinate of the base station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite when satellite j transmits signal data.

[0020] As a preferred solution of the present invention, step S102 further includes:

[0021] Based on the pseudorange correction equation and the flow pseudorange observation equation, a pseudorange positioning equation for roughly determining the three-dimensional coordinates of the mobile station is constructed. The pseudorange positioning equation is:

[0022]

[0023] Among them, M represents the mobile station, j represents the satellite, is the pseudorange observation value from the mobile station to satellite j, is the geometric distance from the mobile station to satellite j, δt M j is the clock difference matrix parameter between the mobile station receiver and the satellite, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the mobile station, is the pseudorange observation noise and other non-modeled errors, (X M ,Y M ,Z M ) is the three-dimensional coordinate of the mobile station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite when satellite j transmits signal data.

[0024] As a preferred solution of the present invention, in step S103, the method for constructing the GNSS positioning function includes:

[0025] Set b = δt M j -δt B j, SV=(X M ,Y M ,Z M ), U=(X j ,Y j ,Z j ), Pr=P M j -Δρ B j , Er=ε M j , Substituting the pseudo-range positioning equation g1, we get:

[0026] Pr=|SV-U|+b+Er;

[0027] Convert Pr = |SV-U| + b + Er into a nonlinear mapping function form, and the GNSS positioning function is obtained as follows:

[0028] SV=F(U,Pr);

[0029] Where F represents the nonlinear mapping function of the GNSS positioning function, SV represents the three-dimensional coordinates of the mobile station, U represents the three-dimensional coordinates of the satellite at the time when satellite j transmits signal data, Pr represents the pseudorange correlation value between the mobile station and the base station, b represents the clock difference correlation value between the mobile station and the base station, and Er represents the pseudorange observation noise and other non-modeled errors.

[0030] As a preferred solution of the present invention, in step S2, the method for converting the GNSS precise positioning model includes:

[0031] Step S201, building a neural network, and converting the GNSS positioning function into a neural network model form based on the neural network;

[0032] Step S202, collecting training samples, and performing model training on the neural network based on the training samples to obtain the GNSS precise positioning model, so as to achieve accurate output of the three-dimensional coordinates of the mobile station by using the pseudorange observation values ​​and the three-dimensional coordinates of all satellites in the GNSS satellite system.

[0033] As a preferred solution of the present invention, in step S201, the method for building the neural network includes:

[0034] Based on the GNSS positioning function SV=F(U,Pr), the input of the neural network is set to U and Pr, and the output is set to SV, wherein the input form of the neural network is:

[0035] {I jt |j∈[1,n],t∈[1,T]},

[0036] in, U t It is represented by the three-dimensional coordinates (X jt , Y jt , Z jt ), Pr t It is characterized by the pseudorange correlation value P of the mobile station and the base station at time t M jt -Δρ B jt , T represents the final value at time t, and n is the total number of satellites in the GNSS satellite system;

[0037] The output form of the neural network is: t |t∈[1,T]},

[0038] in, SV t It is represented by the three-dimensional coordinates (X Mt , Y Mt , Z Mt ), T represents the final value at time t;

[0039] The network body of the neural network is set to an ELM neural network, the number of hidden layer neurons is set to 8500, the neuron activation function is set to a sin function, and the loss function is set to a root mean square error function.

[0040] As a preferred solution of the present invention, in step S202, the method for collecting training samples includes:

[0041] A flow grid is evenly set in the target area, and the reference station and the mobile station are distributed in the flow grid to obtain the three-dimensional coordinates SV of the mobile station within a time interval. t , the three-dimensional coordinates of the satellite when satellite j transmits signal data U t , and the pseudorange correlation value Pr of the mobile station and the base station t As a training sample with time series attributes [I jt ,O t ].

[0042] As a preferred solution of the present invention, the training sample [I jt , O t ] is substituted into the neural network for model training to obtain the GNSS precise positioning model.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention utilizes a GNSS combined positioning system that integrates the GPS satellite system, the BDS satellite system, the GLONASS satellite system, and the Galileo satellite system to obtain more available satellites, improve the spatial geometric structure of the satellites, and thus improve the availability, reliability, and accuracy of the GNSS satellite system. The present invention performs pseudo-range differential positioning on the GNSS satellite system to obtain a positioning function for the position of the GNSS receiver, and uses a neural network to perform nonlinear mapping on the positioning function to obtain a GNSS precise positioning model that fully matches the mapping relationship between the satellite positioning data. The positioning operation has high efficiency and high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0046] Figure 1 A flow chart of a method for enhancing GNSS positioning accuracy provided by an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of the distribution structure of reference stations and mobile stations in a mobile grid provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] like Figure 1As shown, both the base station and the mobile station are provided with GNSS receivers, and the GNSS receivers in the base station and the mobile station simultaneously receive satellite signals of all satellites of the GNSS satellite system composed of satellite signal data of the GPS satellite system, the BDS satellite system, the GLONASS satellite system and the Galileo satellite system, and simultaneously integrate the GPS satellite system, the BDS satellite system, the GLONASS satellite system and the Galileo satellite system, so that more available satellites can be obtained for positioning, and a larger satellite space structure can be constructed, so that the coordinates of the GNSS receiver position obtained based on the large satellite space structure are more accurate, and the linear solution function used in the positioning solution of the GNSS satellite system cannot completely match the mapping relationship between the positioning data, so the present invention provides a GNSS positioning accuracy enhancement method, which uses a neural network to perform a large amount of data training to obtain a GNSS precise positioning model that completely matches the mapping relationship between the positioning data, thereby improving the positioning solution accuracy.

[0050] The present invention provides a method for enhancing GNSS positioning accuracy, comprising the following steps:

[0051] Step S1, constructing a GNSS positioning function based on the mobile station, the base station and the GNSS satellite system;

[0052] In step S1, the method for constructing the GNSS positioning function includes:

[0053] Step S101: construct a reference pseudorange observation equation based on the reference station and the GNSS satellite system. The reference pseudorange observation equation is characterized by pseudorange observation values ​​from the reference station to all satellites in the GNSS satellite system. The reference pseudorange observation equation is:

[0054]

[0055] Among them, B represents the base station, j represents the satellite, is the pseudorange observation value from the reference station to satellite j, is the geometric distance from the reference station to satellite j, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the reference station, is the pseudorange observation noise and other non-modeled errors, (X B ,Y B ,Z B ) is the three-dimensional coordinate of the base station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite at the moment when satellite j transmits signal data;

[0056] Step S101 also includes correcting the reference pseudorange observation equation to obtain a pseudorange correction equation, where the pseudorange correction equation is represented by a pseudorange correction value from the reference station to all satellites in the GNSS satellite system, wherein the pseudorange correction equation is:

[0057] Δρ B j =P B j -ρ B j =δt B j +I B j +T B j +ε B j ;

[0058] Among them, B represents the base station, j represents the satellite, is the pseudorange observation value from the reference station to satellite j, is the geometric distance from the reference station to satellite j, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the reference station, is the pseudorange observation noise and other non-modeled errors, (X B ,Y B ,Z B ) is the three-dimensional coordinate of the base station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite when satellite j transmits signal data.

[0059] Step S102: construct a mobile pseudorange observation equation based on the mobile station and the GNSS satellite system. The reference pseudorange observation equation is characterized by pseudorange observation values ​​from the mobile station to all satellites in the GNSS satellite system. The mobile pseudorange observation equation is:

[0060] f2=P M j =ρ M j +δt M j +I M j +T M j +ε M j ;

[0061] Among them, M represents the mobile station, j represents the satellite, is the pseudorange observation value from the mobile station to satellite j, is the geometric distance from the mobile station to satellite j, δt M j is the clock difference matrix parameter between the mobile station receiver and the satellite, are the ionospheric error and tropospheric error of the mobile station, is the pseudorange observation noise and other non-modeled errors;

[0062] Step S102 also includes:

[0063] The ionospheric error and tropospheric error of the reference station and the mobile station have spatial correlation. In the process of constructing the first positioning equation, the tropospheric delay and ionospheric delay of the reference station and the mobile station are kept equal, that is,

[0064] Based on the pseudorange correction equation and the flow pseudorange observation equation, a pseudorange positioning equation is constructed to roughly determine the three-dimensional coordinates of the mobile station. The pseudorange positioning equation is:

[0065]

[0066] Among them, M represents the mobile station, j represents the satellite, is the pseudorange observation value from the mobile station to satellite j, is the geometric distance from the mobile station to satellite j, δt M j is the clock difference matrix parameter between the mobile station receiver and the satellite, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the mobile station, is the pseudorange observation noise and other non-modeled errors, (X M ,YM,Z M ) is the three-dimensional coordinate of the mobile station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite when satellite j transmits signal data.

[0067] The above steps effectively quantify the deviations between systems, which can effectively reduce common errors, increase the amount of observations, and improve the strength and performance of parameter solution. By constructing the GNSS positioning function through the principle of system difference, some common errors such as satellite clock error, receiver clock error of base station and mobile station, model deviation, etc. can be eliminated, which ultimately improves the model strength of parameter solution and parameter estimation performance to achieve precise positioning.

[0068] Step S103: construct a GNSS positioning function based on the reference pseudorange observation equation and the mobile pseudorange observation equation. The GNSS positioning function is characterized by a nonlinear mapping relationship between the three-dimensional coordinates of the mobile station and the pseudorange observation values ​​and the three-dimensional coordinates of all satellites in the GNSS satellite system.

[0069] In step S103, the method for constructing the GNSS positioning function includes:

[0070] Set b = δt M j -δt B j , SV=(X m ,Y m ,Z m ), U=(X j ,Y j ,Z j ), Pr=P M j -Δρ B j , Er=ε M j , replace the pseudo-range positioning equation g1, and get:

[0071] Pr=|SV-U|+b+Er;

[0072] Convert Pr = |SV-U| + b + Er into a nonlinear mapping function form, and the GNSS positioning function is obtained as follows:

[0073] SV=F(U,Pr);

[0074] Where F represents the nonlinear mapping function of the GNSS positioning function, SV represents the three-dimensional coordinates of the mobile station, U represents the three-dimensional coordinates of the satellite at the time when satellite j transmits signal data, Pr represents the pseudorange correlation value between the mobile station and the base station, b represents the clock difference correlation value between the mobile station and the base station, and Er represents the pseudorange observation noise and other non-modeled errors.

[0075] Pr=|SV-U|+b+Er is a continuous and differentiable function, and there is no closed-form solution. The neural network can infinitely approximate the continuous differentiable function under the condition of sufficient training samples. Therefore, the Pr=|SV-U|+b+Er function can be converted into a nonlinear mapping function form SV=F(U,Pr) that infinitely approximates the mapping relationship between positioning data. That is, the data relationship between SV, U, and Pr can be quantified as a nonlinear mapping with a fixed form of input and output. F is a nonlinear mapping function body of the GNSS positioning function. It has no specific mathematical expression, but it is continuously differentiable. Using a neural network to perform positioning solution based on SV=F(U,Pr) can obtain higher positioning solution accuracy than solving Pr=|SV-U|+b+Er.

[0076] Step S2: Use a neural network to map and approximate the GNSS positioning function, and convert the GNSS positioning function into a GNSS precise positioning model to enhance the GNSS positioning accuracy.

[0077] In step S2, the conversion method of the GNSS precise positioning model includes:

[0078] Step S201, building a neural network, and converting the GNSS positioning function into a neural network model form based on the neural network;

[0079] Step S202: Collect training samples, and train the neural network model based on the training samples to obtain a GNSS precise positioning model, so as to accurately output the three-dimensional coordinates of the mobile station using pseudo-range observations and the three-dimensional coordinates of all satellites in the GNSS satellite system.

[0080] In step S201, the method for building a neural network includes:

[0081] Based on the GNSS positioning function SV = F (U, Pr), the input of the neural network is set to U and Pr, and the output is set to SV, where the input form of the neural network is:

[0082] {I jt |j∈[1,n],t∈[1,T]},

[0083] in, U t It is represented by the three-dimensional coordinates (X jt , Y jt , Z jt ), Pr t It is characterized by the pseudorange correlation value P of the mobile station and the base station at time t M jt -Δρ B jt, T represents the final value at time t, and n is the total number of satellites in the GNSS satellite system;

[0084] The output form of the neural network is: t |t∈[1,T]},

[0085] in, SV t It is represented by the three-dimensional coordinates (X Mt ,Y Mt ,Z Mt ), T represents the final value at time t;

[0086] The network body of the neural network is set to the ELM neural network, the number of hidden layer neurons is set to 8500, the neuron activation function is set to the sin function, and the loss function is set to the root mean square error function.

[0087] The ELM neural network used in this embodiment may also use a BP neural network, a LS network, etc., all of which are within the protection scope of this embodiment.

[0088] Based on the principle of neural networks, as long as there are enough hidden layer nodes in the neural network, the nonlinear mapping F can be infinitely approximated. Therefore, the neural network method is used to train the neural network to approximate the nonlinear mapping F, instead of solving the equation Pr=|SV-U|+b+Er to obtain the three-dimensional coordinates of the mobile station. It is only necessary to input the observation values ​​into the GNSS precise positioning model trained by the neural network during the positioning solution to obtain the three-dimensional coordinates of the mobile station.

[0089] like Figure 2 As shown, in step S202, the method for collecting training samples includes:

[0090] Evenly set the flow grid in the target area, and distribute the base station and the mobile station in the flow grid to obtain the three-dimensional coordinates SV of the mobile station within a time interval t , the three-dimensional coordinates of the satellite when satellite j transmits signal data U t , and the pseudorange correlation value Pr of the mobile station and the base station t As a training sample with time series attributes [I jt ,O t ].

[0091] The training samples [I jt ,O t ] is substituted into the neural network for model training to obtain the GNSS precise positioning model.

[0092] In this embodiment, there is no limit on the number of base stations and mobile stations set in the mobile network, which can be customized by the user. As the number of base stations and mobile stations increases, the accuracy of the GNSS precise positioning model will be improved, but at the same time, the structural scale of the neural network will be required to be higher. Therefore, a trade-off between accuracy and efficiency needs to be made when choosing.

[0093] The present invention utilizes a GNSS combined positioning system that integrates the GPS satellite system, the BDS satellite system, the GLONASS satellite system, and the Galileo satellite system to obtain more available satellites, improve the spatial geometric structure of the satellites, and thus improve the availability, reliability, and accuracy of the GNSS satellite system. The present invention performs pseudo-range differential positioning on the GNSS satellite system to obtain a positioning function for the position of the GNSS receiver, and uses a neural network to perform nonlinear mapping on the positioning function to obtain a GNSS precise positioning model that fully matches the mapping relationship between the satellite positioning data. The positioning operation has high efficiency and high accuracy.

[0094] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for enhancing GNSS positioning accuracy, characterized in that: The following steps are involved: Step S1, constructing a GNSS positioning function based on the mobile station, the base station and the GNSS satellite system; Step S2, mapping and approximating the GNSS positioning function using a neural network, converting the GNSS positioning function into a GNSS precise positioning model to enhance the GNSS positioning accuracy; in the step S1, the method for constructing the GNSS positioning function includes: Step S101: construct a reference pseudorange observation equation based on a reference station and a GNSS satellite system, wherein the reference pseudorange observation equation is characterized by pseudorange observation values ​​from the reference station to all satellites in the GNSS satellite system, wherein the reference pseudorange observation equation is: Among them, B represents the base station, j represents the satellite, is the pseudorange observation value from the reference station to satellite j, is the geometric distance from the reference station to satellite j, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the reference station, is the pseudorange observation noise and other non-modeled errors, (X B ,Y B ,Z B ) is the three-dimensional coordinate of the base station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite at the moment when satellite j transmits signal data; Step S102: construct a mobile pseudorange observation equation based on the mobile station and the GNSS satellite system, wherein the reference pseudorange observation equation is characterized by pseudorange observation values ​​from the mobile station to all satellites in the GNSS satellite system, wherein the mobile pseudorange observation equation is: f2=P M j =ρ M j +δt M j +I M j +T M j +ε M j ; Among them, M represents the mobile station, j represents the satellite, is the pseudorange observation value from the mobile station to satellite j, is the geometric distance from the mobile station to satellite j, δt M j is the clock difference matrix parameter between the mobile station receiver and the satellite, are the ionospheric error and tropospheric error of the mobile station, is the pseudorange observation noise and other non-modeled errors; Step S103, constructing a GNSS positioning function based on the reference pseudorange observation equation and the mobile pseudorange observation equation, wherein the GNSS positioning function is characterized by a nonlinear mapping relationship between the three-dimensional coordinates of the mobile station and the pseudorange observation values ​​and the three-dimensional coordinates of all satellites in the GNSS satellite system; Step S102 also includes: Based on the pseudorange correction equation and the flow pseudorange observation equation, a pseudorange positioning equation is constructed to roughly determine the three-dimensional coordinates of the mobile station. The pseudorange positioning equation is: Among them, M represents the mobile station, j represents the satellite, is the pseudorange observation value from the mobile station to satellite j, is the geometric distance from the mobile station to satellite j, δt M j is the clock difference matrix parameter between the mobile station receiver and the satellite, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the mobile station, is the pseudorange observation noise and other non-modeled errors, (X M ,Y M ,Z M ) is the three-dimensional coordinate of the mobile station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite at the moment when satellite j transmits signal data; In step S103, the method for constructing the GNSS positioning function includes: Set b = δt M j -δt B j , SV=(X M ,Y M ,Z M ), U=(X j ,Y j ,Z j ), Pr=P M j -Δρ B j , Er=ε M j , replace the pseudo-range positioning equation g1, and get: Pr=|SV-U|+b+Er; Convert Pr = |SV-U| + b + Er into a nonlinear mapping function form, and the GNSS positioning function is obtained as follows: SV=F(U,Pr); Where, F represents the nonlinear mapping function of the GNSS positioning function, SV represents the three-dimensional coordinates of the mobile station, U represents the three-dimensional coordinates of the satellite at the time when satellite j transmits the signal data, Pr represents the pseudorange correlation value between the mobile station and the base station, b represents the clock difference correlation value between the mobile station and the base station, and Er represents the pseudorange observation noise and other non-modeled errors; Step S201, building a neural network, and converting the GNSS positioning function into a neural network model form based on the neural network; Step S202, collecting training samples, and training the neural network model based on the training samples to obtain a GNSS precise positioning model, so as to accurately output the three-dimensional coordinates of the mobile station using the pseudo-range observation values ​​and the three-dimensional coordinates of all satellites in the GNSS satellite system; In step S201, the method for building the neural network includes: Based on the GNSS positioning function SV = F (U, Pr), the input of the neural network is set to U and Pr, and the output is set to SV, where the input form of the neural network is: {I jt |j∈[1,n],t∈[1,T]}, in, U t It is represented by the three-dimensional coordinates (X jt ,Y jt ,Z jt ), Pr t It is characterized by the pseudorange correlation value P of the mobile station and the base station at time t M jt -Δρ B jt , T represents the final value at time t, and n is the total number of satellites in the GNSS satellite system; The output form of the neural network is: t |t∈[1,T]}, in, SV t It is represented by the three-dimensional coordinates (X Mt ,Y Mt ,Z Mt ), T represents the final value at time t; The network body of the neural network is set to ELM neural network, the number of hidden layer neurons is set to 8500, the neuron activation function is set to sin function, and the loss function is set to root mean square error function; The step S101 further includes correcting the reference pseudorange observation equation to obtain a pseudorange correction equation, wherein the pseudorange correction equation is characterized by a pseudorange correction value from the reference station to all satellites in the GNSS satellite system, wherein the pseudorange correction equation is: Dr. B j =P B j -r B j =δt B j +I B j +T B j +e B j ; Among them, B represents the base station, j represents the satellite, is the pseudorange observation value from the reference station to satellite j, is the geometric distance from the reference station to satellite j, δt B j is the clock difference matrix parameter between the base station receiver and the satellite, are the ionospheric error and tropospheric error of the reference station, is the pseudorange observation noise and other non-modeled errors, (X B ,Y B ,Z B ) is the three-dimensional coordinate of the base station, (X j ,Y j ,Z j ) is the three-dimensional coordinate of the satellite at the moment when satellite j transmits signal data; In step S202, the method for collecting training samples includes: A flow grid is evenly set in the target area, and the reference station and the mobile station are distributed in the flow grid to obtain the three-dimensional coordinates SV of the mobile station within a time interval. t , the three-dimensional coordinates of the satellite when satellite j transmits signal data U t , and the pseudorange correlation value Pr of the mobile station and the base station t As a training sample with time series attributes [I jt ,O t ]; The training samples [I jt ,O t ] is substituted into the neural network for model training to obtain the GNSS precise positioning model.

Images