A data-driven single-satellite timing estimation error correction method based on transfer learning

Abstract

A data-driven single-satellite timing estimation error correction method based on transfer learning relates to the field of satellite communication timing technology, including constructing a transfer learning network and a transfer perception network according to a data-driven paradigm; using a ground base station to receive a forwarding single-satellite signal, obtaining a code pseudorange state observation vector and a carrier pseudorange state observation vector, and generating a carrier integer fuzzy state vector; according to the carrier integer fuzzy state vector, using the decision of the transfer perception network, controlling the transfer learning network to output a timing deviation estimation value; the timing deviation estimation accuracy can be improved, while reducing the complexity of modeling.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite communication timing, and in particular to a data-driven single-satellite timing estimation error correction method based on transfer learning. Background Art

[0002] The success of the China area positioning system (CAPS) provides key support for the "integrated air-space-ground-sea" architecture of the future sixth generation (6G) communications. The accuracy of its timing estimation directly affects the ubiquitous communication application scenarios in the 6G architecture.

[0003] In the CAPS-based single-satellite forwarding timing system, it is usually necessary to perform complex mathematical modeling on the temporal / spatial variation characteristics (such as the troposphere, relativistic effects, etc.) in the signal propagation path, including topographic changes and atmospheric characteristics, to minimize the impact of error terms in timing bias estimation. However, mathematical modeling based on observations is not only complex, but also difficult to adapt to rapidly changing temporal / spatial characteristics. In addition, the devices used to collect observations often have unavoidable nonlinear characteristics that are difficult to obtain through linear models, and nonlinear modeling is difficult. For example, the Kalman filter-based timing bias prediction and estimation error correction algorithm requires the establishment of a linear Gaussian white noise model, so it is not suitable for hardware devices with nonlinear distortion, which will lead to serious prediction errors and estimation error correction failure.

[0004] A Chinese patent with publication number CN117880955A discloses a pseudo-satellite time synchronization method and system based on error repair. In terms of the overall solution, the method addresses the problem of high cost of traditional pseudo-satellite time synchronization from the perspective of algorithm rather than hardware, and adopts a deep learning model to predict and reduce the receiver clock error, thereby improving the accuracy of pseudo-satellite time synchronization under the premise of controllable costs; however, the method still uses a mathematical model method, which leads to high computational complexity and long computational time. Summary of the invention

[0005] The purpose of the present invention is to provide a data-driven single-satellite timing estimation error correction method based on transfer learning, which can improve the timing deviation estimation accuracy while reducing the complexity of modeling.

[0006] The present invention is achieved through the following technical solutions:

[0007] A data-driven single-satellite timing estimation error correction method based on transfer learning, comprising:

[0008] S1. Based on the data-driven paradigm, build a transfer learning network and a transfer perception network;

[0009] S2. Use a ground base station to receive a forwarding single-satellite signal, obtain a code pseudorange state observation vector and a carrier pseudorange state observation vector, and generate a carrier integer ambiguity state vector;

[0010] S3. Based on the carrier integer fuzzy state vector, the decision of the transfer perception network is used to control the transfer learning network to output the timing deviation estimate.

[0011] Furthermore, the transfer learning network and the transfer perception network include an initial layer, a hidden layer 1, a hidden layer 2, a semantic layer and an output layer.

[0012] Furthermore, the steps of constructing the transfer learning network and the transfer perception network are specifically as follows:

[0013] S11. Design the initial layer to extract common features, expressed as:

[0014] ;

[0015] Said It is an abstract domain, which contains the characteristics of the target domain / source domain; is the carrier integer ambiguity state vector; It is the abstract source domain feature;

[0016] S12. Design hidden layer 1 extraction Local features in the domain , expressed as:

[0017] ;

[0018] Said is the abstraction function of hidden layer 1; is the network weight parameter of hidden layer 1;

[0019] S13. Design hidden layer 2 extraction Domain texture features in , expressed as:

[0020] ;

[0021] Said is the abstraction function of hidden layer 2; is the network weight parameter of hidden layer 2;

[0022] S14. Design semantic layer extraction Domain implicit knowledge in , expressed as:

[0023] ;

[0024] Said It is the abstraction function of the semantic layer; is the network weight parameter of the semantic layer;

[0025] S15. Design a migration-aware network as the output layer and make decisions , expressed as:

[0026] ;

[0027] Said It is the abstract function of the output layer; are the transfer-aware network weight parameters for the output layer.

[0028] Furthermore, the carrier integer ambiguity state vector is obtained by the following process:

[0029] S21. Obtain the code pseudo-range state observation quantity of the ground base station measurement signal from the ground station through the satellite to the receiver :

[0030] ;

[0031] Said The time when the ground base station observes the received signal; Obtained by deciphering the telegram; is the information bit period; is the number of bits received at the current observation time; is the chip period; is the number of chips received at the current observation time; is the code phase; is the digital frequency synthesis constant; is the constant speed of light;

[0032] S22. Obtain the carrier pseudo-range state observation value of the ground base station measurement signal from the ground station through the satellite to the receiver:

[0033] ;

[0034] Said For ground base stations, measure the phase difference of the signal from the ground station through the satellite to the receiver; is the RF carrier frequency;

[0035] S23. Construct a code pseudorange state observation vector according to the code pseudorange state observation quantity , according to the carrier pseudorange state observation, construct the carrier pseudorange state observation vector ,Right now:

[0036] ;

[0037] Among them, the is the length of the observation vector;

[0038] S24. According to and , calculate the carrier integer ambiguity state vector.

[0039] Furthermore, the carrier integer ambiguity state vector , obtained by the following calculation formula, namely:

[0040] ;

[0041] .

[0042] Further, the timing deviation estimate is calculated by the following process:

[0043] S31. Based on the carrier integer fuzzy state vector , and get the decision output of the transfer learning network ;

[0044] S32. Generate the satellite position state observation value of the current period through the satellite ephemeris iteration solution algorithm;

[0045] S33. Obtaining the satellite-to-ground distance observation according to the satellite position state observation;

[0046] S34. Based on the satellite-to-ground distance observation, the carrier pseudorange state observation vector and transfer learning network decision output , calculate the timing bias estimate .

[0047] Furthermore, the S32 satellite position state observation The calculation method is as follows:

[0048] ;

[0049] ;

[0050] The satellite ephemeris is obtained by de-framing the navigation message; is the Chebyshev polynomial fitting coefficient; It is the satellite ephemeris update period; is the observation interval; is the satellite ephemeris update time; is the iterative fitting parameter, Planning time The i-order fitted value of .

[0051] Furthermore, the calculation formula of the satellite-to-earth distance observation is specifically:

[0052] ;

[0053] Said It is the rectangular coordinate position of the ground receiver, measured by high-precision instruments on site.

[0054] Furthermore, the calculation formula for estimating the timing deviation is specifically:

[0055] .

[0056] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0057] The present invention discloses a data-driven single-satellite timing estimation error correction method based on transfer learning. The method is constructed according to deep learning and forwarding single-satellite timing principles, and can improve the timing deviation estimation accuracy under the influence of device nonlinearity while avoiding complex mathematical modeling.

[0058] And this method specifically includes constructing a transfer learning network, which is used as a universal function approximator, and capturing and learning implicit data semantic features through a multi-layer network of nonlinear feature transformation to make up for the defects of the physical layer, thereby improving the accuracy of timing deviation estimation.

[0059] In addition, for the scenario of forwarding single-satellite timing requirements, this method can use the transfer learning network to replace the mathematical modeling process of time / space variation characteristics and device nonlinear distortion on the basis of the carrier and code pseudorange observation equations, thereby improving the accuracy of timing deviation estimation while reducing modeling complexity. This brings many implementation plans for future 6GHz or even terahertz satellite communication single-satellite timing, which is of great significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of a method flow of the present invention;

[0061] Figure 2 This is a schematic diagram of the transfer learning network architecture of the present invention;

[0062] Figure 3 A schematic diagram for generating input data for the transfer learning network of the present invention;

[0063] Figure 4 This is a flow chart of the timing deviation estimation of the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0065] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0066] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0067] In the description of the present invention, it should be noted that if the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside" and the like appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the accompanying drawings, or is the orientation or position relationship in which the product of the application is usually placed when used. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0068] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0069] Example 1

[0070] As attached Figure 1 A data-driven single-satellite timing estimation error correction method based on transfer learning is shown, comprising:

[0071] S1. According to the data-driven paradigm, a transfer learning network and a transfer perception network are constructed, including an initial layer, a hidden layer 1, a hidden layer 2, a semantic layer, and an output layer;

[0072] The transfer learning network is a universal function approximator that can capture complex patterns in the input data through training, and the transfer perception network is responsible for decision-making and controlling the output of the transfer learning network; these networks are designed to extract common features, semantic features of domain implicit knowledge, and ultimately output timing deviation estimates;

[0073] S2. The high-precision ground base station receives the forwarding single-satellite signal, obtains the code pseudorange state observation vector and the carrier pseudorange state observation vector, and generates the carrier integer ambiguity state vector; the pseudocode correlation principle is used to measure the time delay of the signal from the system master clock, through the uplink, the satellite itself, the downlink, until it reaches the ground station;

[0074] S3. According to the carrier integer fuzzy state vector, the transfer perception network decides to control the transfer learning network to output the timing deviation estimate; this process involves inputting the carrier integer fuzzy state vector into the transfer learning network, propagating through the network, and finally obtaining the adjusted timing deviation estimate;

[0075] The pseudorange state observation vector and carrier pseudorange state observation vector obtained by the high-precision ground base station can be measured according to the existing pseudocode correlation principle to measure the pseudorange of the signal from the system master clock, through the uplink space, the satellite, and the downlink space to the satellite (such as reference documents "Shi Huli, Pei Jun. Solution of the observation equation of the forwarding satellite navigation and positioning system [J]. Chinese Science, 2008 (12): 1687-1701.", "Shi Huli, Sun Xiyan, Li Zhigang. Principle of forwarding satellite navigation [M]. Science Press, 2009."); the transfer learning network can be implemented according to the neural network structure for initializing weights, biases and activation functions in the prior art (such as reference documents "Zhuang Fuzhen, Zhu Yongchun, Zhu Hengshu, Xiong Hui. Transfer learning algorithm: application and time [M]. Machinery Industry Press, 2023.);

[0076] The design method of the transfer learning network and the transfer perception network described in step S1 can be further described in the attached Figure 2 As shown, specifically including:

[0077] Extract the common features of the transfer learning network input, expressed as:

[0078] ;

[0079] Said It is an abstract domain, which contains the characteristics of the target domain / source domain; is the carrier integer fuzzy state vector, but when the transfer learning network is initially constructed, For the abstract target domain, in the subsequent S3 calculation, the carrier integer ambiguity state vector calculated by S2 is Substitute here; It is an abstract source domain feature, collected by measurement equipment in the past period of time;

[0080] extract The local features of the domain in the hidden layer 1 are designed for feature extraction, which can be expressed as:

[0081] ;

[0082] Said is the abstraction function of hidden layer 1; is the network weight parameter of hidden layer 1;

[0083] extract In the domain texture feature, hidden layer 2 is designed for feature extraction, which is expressed as:

[0084] ;

[0085] Said is the abstraction function of hidden layer 2; is the network weight parameter of hidden layer 2;

[0086] To improve the decision-making performance of the transfer learning network, the semantic layer is designed to extract semantic features with domain implicit knowledge, namely:

[0087] ;

[0088] Said It is the abstraction function of the semantic layer; is the network weight parameter of the semantic layer;

[0089] To determine the output of the transfer learning network, the transfer-aware network is designed to complete the following equation:

[0090] ;

[0091] Said It is the abstract function of the output layer; The weight parameters of the transfer-aware network used for the output layer are determined by the implicit knowledge obtained from past sample feature training collected by the measurement device.

[0092] The carrier integer ambiguity state vector described in step S2 can be further described as follows: Figure 3 As shown, specifically including:

[0093] According to the received signal from the ground station to the receiver via the satellite, the high-precision ground base station measures the code pseudo-range state observation, namely:

[0094] ;

[0095] Said The time when the high-precision ground base station observes the received signal; Obtained by deciphering the telegram; It is the information bit period, which is determined by the protocol standard; is the number of bits received at the current observation time; It is the chip period, which is determined by the protocol standard; is the number of chips received at the current observation time; is the code phase, measured by field equipment and implementation devices; is the digital frequency synthesis constant, which is determined by the hardware indicators; is the constant speed of light, 299792458m / s.

[0096] According to the received signal from the ground station to the receiver via the satellite, the high-precision ground base station measures the carrier pseudorange state observation, that is, , For ground base stations, measure the phase difference of the signal from the ground station through the satellite to the receiver; is the RF carrier frequency.

[0097] according to Secondary measurement and , construct code pseudorange state observation vector and carrier pseudorange state observation vector ,Right now:

[0098] ;

[0099] Said It is the length of the observation vector, which is determined by the number of observations reported per second as required by the engineering indicator.

[0100] To generate the carrier integer fuzzy state vector for transfer learning network input , there are the following processes:

[0101] ;

[0102] ;

[0103] Said As input to the transfer learning network.

[0104] The timing deviation estimation in step S3 can be further described as follows: Figure 4 As shown, specifically including:

[0105] According to the carrier integer ambiguity state vector described in step 2 , and get the decision output of the transfer learning network ;

[0106] Through the satellite ephemeris iteration solution algorithm, the satellite position state observation quantity of the current period The calculation method is as follows:

[0107] ;

[0108] ;

[0109] The satellite ephemeris is obtained by de-framing the navigation message; are the Chebyshev polynomial fitting coefficients, obtained from long-term observations by high-precision ground base stations; It is the satellite ephemeris update period, which is determined by the protocol standard; It is the observation interval, which is determined by the protocol standard; is the satellite ephemeris update time, which is calculated from the telegram information; Iterative fitting parameters are set by engineering experience. Planning time The i-order fitting value of ;

[0110] According to the satellite position state observation , and obtain the observed distance between the satellite and the earth, namely:

[0111] ;

[0112] Said It is the rectangular coordinate position of the ground receiver, measured by high-precision instruments on site.

[0113] According to the distance between the satellite and the earth , carrier pseudorange state observation vector and transfer learning network decision output , complete the timing deviation estimation , expressed as:

[0114] .

[0115] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data-driven single-satellite timing estimation error correction method based on transfer learning, characterized in that: include: S1. Based on the data-driven paradigm, build a transfer learning network and a transfer perception network; The transfer learning network and the transfer perception network include an initial layer, a hidden layer 1, a hidden layer 2, a semantic layer and an output layer; The steps for constructing the transfer learning network and the transfer perception network are specifically as follows: S11. Design the initial layer to extract common features, expressed as: ； Said It is an abstract domain, which contains the characteristics of the target domain / source domain; is the carrier integer ambiguity state vector; It is the abstract source domain feature; S12. Design hidden layer 1 extraction Local features in the domain , expressed as: ； Said is the abstraction function of hidden layer 1; is the network weight parameter of hidden layer 1; S13. Design hidden layer 2 extraction Domain texture features in , expressed as: ； Said is the abstraction function of hidden layer 2; is the network weight parameter of hidden layer 2; S14. Design semantic layer extraction Domain implicit knowledge in , expressed as: ； Said It is the abstraction function of the semantic layer; is the network weight parameter of the semantic layer; S15. Design a migration-aware network as the output layer and make decisions , expressed as: ； Said It is the abstract function of the output layer; is the transfer-aware network weight parameter for the output layer; S2. Use a ground base station to receive a forwarding single-satellite signal, obtain a code pseudorange state observation vector and a carrier pseudorange state observation vector, and generate a carrier integer ambiguity state vector; S21. Obtain the code pseudo-range state observation quantity of the ground base station measurement signal from the ground station through the satellite to the receiver : ； Said The time when the ground base station observes the received signal; Obtained by deciphering the telegram; is the information bit period; is the number of bits received at the current observation time; is the chip period; is the number of chips received at the current observation time; is the code phase; is the digital frequency synthesis constant; is the constant speed of light; S22. Obtain the carrier pseudo-range state observation value of the ground base station measurement signal from the ground station through the satellite to the receiver: ； Said For ground base stations, measure the phase difference of the signal from the ground station through the satellite to the receiver; is the RF carrier frequency; S23. Construct a code pseudorange state observation vector according to the code pseudorange state observation quantity , according to the carrier pseudorange state observation, construct the carrier pseudorange state observation vector ,Right now: ； Among them, the is the length of the observation vector; S24. According to and , calculate the carrier integer ambiguity state vector; The carrier integer ambiguity state vector , obtained by the following calculation formula, namely: ； ； S3. Based on the carrier integer fuzzy state vector, the decision of the migration perception network is used to control the migration learning network to output the timing deviation estimate; The timing deviation estimate is calculated by the following process: S31. According to the carrier integer fuzzy state vector, the transfer learning network decision output is obtained; S32. Generate the satellite position state observation value of the current period through the satellite ephemeris iteration solution algorithm; The S32 satellite position state observation The calculation method is as follows: ； ； The satellite ephemeris is obtained by de-framing the navigation message; is the Chebyshev polynomial fitting coefficient; It is the satellite ephemeris update period; is the observation interval; It is the satellite ephemeris update time; is the iterative fitting parameter, Planning time The i-order fitting value of ; S33. Obtaining the satellite-to-ground distance observation according to the satellite position state observation; The satellite-to-earth distance observation The calculation formula is as follows: ； Said It is the rectangular coordinate position of the ground receiver, which is measured by high-precision instruments on site; S34. Calculate the timing deviation estimate based on the satellite-to-ground distance observation, the carrier pseudorange state observation vector and the transfer learning network decision output; The timing bias estimate The calculation formula is as follows: 。

Images