Satellite positioning enhancement device, correction information generation method, computer program

Abstract

Generate ionospheric delay correction information for satellite positioning users with high precision for one cycle. 【Solution】Based on the output result of a total electron number prediction model that is trained to output a predicted value of the total electron number at a future time (e.g., one hour later) when data such as atmospheric environmental characteristics and solar altitude angle data are input, a correction parameter (a new correction parameter replacing the ionospheric correction parameter of the existing Klobuchar model) ECP for correcting the ionospheric error is generated using an electron density distribution model expanded over a wide area by a spherical harmonic function with the total electron number at the ionospheric penetration point of the received date and time calculated using observation data obtained from ranging signals of multiple frequencies (e.g., L1 wave and L2 wave).

Description

Technical Field

[0001] The present invention relates to satellite positioning enhancement technology, for example, a correction technology for ionospheric errors for high-precision independent positioning in a single-frequency positioning terminal.

Background Art

[0002] An inevitable problem in the Global Navigation Satellite System (GNSS) is the influence on radio wave propagation in the ionosphere that occurs irregularly over the Earth. The ionosphere is a region where electrons are ionized due to solar activity, geomagnetic activity, etc. The electron density distribution in the ionosphere strongly depends on the altitude from the Earth's surface, time of day, and season (tilt of the Earth), refracts the radio wave propagation path from the positioning satellite to the receiver, and causes signal delay. Such signal delay is called "ionospheric delay", the delay time is called "ionospheric delay amount", and the ionospheric delay error based on it is called "ionospheric error". In GNSS, accurately grasping the electron density distribution and its variation status in the ionosphere at the positioning time and correcting the ionospheric error are important for improving positioning accuracy.

[0003] Regarding the correction of ionospheric errors, Non-Patent Document 1 discloses a technique for calculating the Total Electron Content (TEC) over Japan using the latest observation data obtained at the electronic reference points of the Geospatial Information Authority of Japan, and then generating ionospheric delay correction parameters of the Klobuchar model adjusted to the specified area. The "Klobuchar model" is a model that approximates by combining a constant value of 5 [ns] for the change in the ionospheric delay amount per day and the upper half of a cosine curve with a maximum at 14:00 local time (local time at the receiver position) (Non-Patent Document 2). According to the Klobuchar model, when the local time is t, the period of the cosine curve is T, the amplitude is A, the unit of time is s, and the speed of light is c, the ionospheric delay amount δI in the zenith direction from the receiving point in the region where the cosine curve is involved (14×3600 - T / 2 < t < 14×3600 + T / 2) Z can be expressed by Equation 1.

[0004] [Number]

[0005] The amplitude A and period T of the cosine curve can be expressed as a cubic polynomial of the magnetic latitude φm as shown in Equation 2 and Equation 3. [Number] [Number]

[0006] In Equation 2 and Equation 3, the eight parameters of (α 0~3 , β 0~3 ) are ionospheric delay correction parameters, which are stored in the navigation message from GPS (Global Positioning System) satellites and distributed worldwide. On the positioning terminal side, the ionospheric delay amount δI 0~3 , β 0~3 ) corresponding to the receiving point coordinates and receiving date and time is calculated from these ionospheric delay correction parameters, and the ionospheric error is corrected. It is said that this correction can generally reduce about half of the ionospheric error. Z

[0007] The update period of the ionospheric delay correction parameters of the Klobuchar model is as long as several days in some cases. To make up for this point, in Japan's QZSS (Quasi-Zenith Satellite System), ionospheric delay correction parameters specialized for two regions in the vicinity of Japan and Asia and Oceania (Figure 2: page 147, section 5.9.3 of Non-Patent Document 3) are generated every hour and distributed from the quasi-zenith satellite, aiming to further reduce the ionospheric error (Non-Patent Document 3).

[0008] ​Various techniques have also been proposed to improve the prediction of the electron density distribution that varies in the ionosphere. For example, in the ionospheric electron density distribution estimation system disclosed in Patent Document 1, a modified Kalman filter, that is, a filter (matrix) extended to non-linear elements, is used to determine the presence or absence of cycle slips, thereby identifying a continuous data collection period. After removing the uncertainty values related to the carrier phase for the identified data collection period, a plurality of carrier phase pseudo-range differences are converted into the number of electrons in the ionosphere, and the electron density distribution of the ionosphere is predicted using this number of electrons in the ionosphere.

Prior Art Documents

Patent Documents

[0009]

Patent Document 1

Non-Patent Documents

[0010]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0011] In the technologies disclosed in Non-Patent Documents 1 to 3, there are cases where the time lag between the time zone when the ionospheric delay correction parameter is generated and the time zone when the ionospheric delay correction parameter is received on the receiver side cannot be ignored. In this case, it cannot be said that the ionospheric delay correction parameter conforms to the state of the ionosphere at the positioning time, and this has a significant impact on the positioning correction accuracy. In addition, since the ionospheric delay correction parameter is generated using the observation data of the electronic reference points (the positions of the observation stations) arranged on land, it is difficult to accurately generate the ionospheric delay correction parameter that can be used in a wide area including land with sparse electronic reference points and the ocean. Such problems also exist similarly in the technology disclosed in Patent Document 1 where the number of observation stations is limited and the processing amount is huge.

[0012] In addition, in the technologies disclosed above, when predicting the electron density distribution, the solar altitude angle and the direction of the sun as seen from the receiving point are not considered. Therefore, the amount of incident sunlight (such as the amount of ultraviolet rays and X-ray radiation) is not reflected, and there remains a problem that the predicted value of the total electron number in the future may deviate significantly from the true value.

[0013] One of the objects of the present invention is to provide a positioning enhancement technology that solves the above problems. Other problems of the present invention will become apparent from the disclosure of this specification.

Means for Solving the Problems

[0014] One aspect of the present invention for achieving the above object is embodied as a computer program. This computer program causes a computer to calculate the total electron content on the radio wave propagation path connecting the positioning satellite and the observation station based on reception signals from one or more positioning satellites observed at the observation station, and to use the total electron content to generate an electron density distribution model of the wide-area ionosphere by a spherical harmonic function or a spherical cap harmonic function; prediction means for inputting the atmospheric environment characteristics including the solar altitude angle at the observation time and the electron density distribution model into a total electron content prediction model constructed by machine learning to derive a predicted value of the total electron content in the wide-area ionosphere at a future time of a predetermined period; correction information generation means for generating correction information for correcting the ionospheric error at the future time of the predetermined period using the predicted value; and is a computer program that operates as such.

Advantages of the Invention

[0015] According to the above aspect, since the total electron content on the radio wave propagation path is calculated based on an electron density distribution model expanded over a wide area using a spherical harmonic function or the like, even for a positioning terminal located in a region with sparse observation stations or on the ocean, the ionospheric error can be accurately corrected. Further, since the predicted value of the total electron content at a future time from the observation time is derived in consideration of the atmospheric environment characteristics including the solar altitude angle, the influence of the time lag is reduced, the predicted value of the total electron content becomes more accurate, and the correction accuracy of the ionospheric error can be improved.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Embodiments for Carrying Out the Invention

[0017] Hereinafter, embodiments in the case where the present invention is applied to a satellite positioning reinforcement device for enhancing high-precision single-code positioning in a positioning terminal will be described. The positioning terminal is a general-purpose information terminal equipped with a receiver and an antenna for receiving satellite signals. In this example, for the sake of convenience, an information terminal equipped with a one-frequency (any one of L1, L2, and L5 frequencies) receiver and an antenna capable of using the Klobuchar model ionospheric delay correction parameters in the above-mentioned GPS or QZSS, that is, an information terminal capable of single-code positioning, will be described.

[0018] [Overall Configuration Example] FIG. 1 is an exemplary diagram of the overall configuration of a satellite positioning system including a satellite positioning reinforcement device according to an embodiment of the present invention. The satellite positioning reinforcement device 1 is configured to include a computer having a server function that can be connected to the Internet, which is an example of a wide-area network N. The satellite positioning reinforcement device 1 can be implemented as a sub-combination such as a GNSS monitoring station that can communicate with a large number of positioning satellites.

[0019] As will be described in detail later, the satellite positioning augmentation device 1 newly generates an original correction parameter ECP that is compatible with the current Klobuchar model's ionospheric delay correction parameters distributed in GPS and QZSS, and replaces the current Klobuchar model's ionospheric delay correction parameters with this, and transmits it toward a plurality of positioning satellites 2n (the "n" attached to the symbol 2 is a natural number of 1 or more, that is, it means that there is one or more. The same applies to other symbols). The positioning satellite 2n can distribute the received ionospheric delay correction parameter ECP to a plurality of positioning terminals 3n using the operating frequencies of GPS or QZSS. The satellite positioning augmentation device 1 can also directly transmit the ionospheric delay correction parameter ECP to each positioning terminal 3n via the wide area network N.

[0020] Note that the correction parameter ECP generated in this embodiment is common to the current Klobuchar model's ionospheric delay correction parameters in GPS and QZSS in that it is based on the Klobuchar model. Therefore, when it is necessary to distinguish between the current Klobuchar model's ionospheric delay correction parameters and the correction parameter ECP generated in this embodiment, the former may be referred to as the "existing correction parameter" and the latter as the "new correction parameter".

[0021] The new correction parameter ECP will be optimized for each region. The existing correction parameters in QZSS are generated by being optimized for two regions, the Japan area A1 and the Asia - Oceania area A2, each indicated by a dashed line in FIG. 2. However, in this embodiment, it is not limited to these regions, and it is also possible to optimize in the same way in other countries or regions.

[0022] The wide area network N is connected to a plurality of information providing systems (including Web systems) that can directly or indirectly provide various information used for satellite positioning to the satellite positioning enhancement device 1. One of the information providing systems is the GNSS integrated data sharing system (MIRAI: Multi-GNSS Integrated Real time and Archived Information system) 4 provided by the Cabinet Office of Japan. MIRAI 4 collects the observation data of GNSS observation stations arranged all over the world in real time, and distributes it together with the reception point coordinates (information such as longitude, latitude, and altitude indicating the position of the antenna of the receiver) of each observation station that has received the observation data. By using the information of MIRAI 4, the satellite positioning enhancement device 1 can acquire and utilize the observation data and navigation messages of a large number of observation stations in real time. Note that the observation station may be called an observation point, a receiver, a receiving antenna, a reception point, or an electronic reference point, and they are synonymous.

[0023] The information providing system includes an information processing device individually provided in each of the individual observation stations 5n for which MIRAI 4 collects and distributes data, or a website operated by an organization that manages the observation stations 5n, and distributes the observation data individually or collectively together with its own identification information and reception point coordinates.

[0024] The information providing system also includes, for example, the websites of domestic and foreign scientific research institutions, in this example, the website of the German national scientific research institution (GFZ: GeoForschungsZentrum Potsdam) 6. From this website, data on at least one of the atmospheric environment characteristics such as solar activity and geomagnetic activity, for example, ultraviolet intensity and geomagnetic intensity, can be acquired in time series and used as the variation amount of the atmospheric environment characteristics. Although not shown in the figure, the information providing system also includes a website that provides a library capable of performing various calculations related to sunlight. For example, by appropriately accessing the well-known "pysolar library" of "Python", the solar altitude angle, solar irradiance, etc. at any point on the earth can be calculated (acquired). The solar altitude angle etc. calculated (acquired) in this way can also be included in and utilized as one of the above-mentioned atmospheric environment characteristics.

[0025] In the wide area network N, in addition to the positioning terminal 3n, information terminals 7n such as personal computers, smartphones, and tablet terminals used by individual users for positioning can also be connected as positioning terminals 3n in a broad sense.

[0026] [Satellite Positioning Enhancement Device] An embodiment example of the satellite positioning enhancement device 1 will be described. FIG. 3 is an exemplary diagram of the hardware configuration of the satellite positioning enhancement device 1. In FIG. 3, the satellite positioning enhancement device 1 includes a server main body 10 and a communication device 11. The server main body 10 is provided with a CPU (Central Processing Unit) 101 that executes various programs including the computer program of the present invention. The CPU 101 is connected via a bus B to a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, a storage 104, a communication device 11, and an interface (I / O) 105 with an input / output device, and controls these hardware devices. The ROM 102 is an example of a non-volatile memory, and stores basic control programs such as BIOS (Basic Input / Output System). The RAM 103 is an example of a volatile memory and is used for temporary data storage and the like. The storage 104 is a storage device having a hard disk drive, a solid state drive, a USB memory, or other recording media. In the storage 104, in addition to an operating system (OS), one or more computer programs for satellite positioning enhancement and a plurality of data files and the like are stored. The computer program is read and executed by the CPU 101. The data stored in the data file is appropriately read by the CPU 101 into the RAM 103 and used for the processing (operation) of the CPU 101.

[0027] The data stored in the data file includes the receiving point coordinates of observation stations around the world, the time series data of the total vertical electron content for each observation station calculated using existing correction parameters and the time series data of predicted values, which are referred to in the process of generating new correction parameters described later, the receiving point coordinates of multiple observation stations, real-time observation data from the observation stations, information on deep learning models, tables for converting various numerical values, and the like.

[0028] The communication device 11 has a wireless communication board 111 and a wired communication board 112. The wireless communication board 111 enables wireless communication with, for example, the positioning satellite 2n. The wireless communication board 111 also enables wireless communication between the wide area network N and the LAN (Local Area Network) via an adapter of a predetermined frequency. The wired communication board 112 enables wired communication with, for example, the above-mentioned positioning terminal 3n, information providing systems (4, 5n, 6), information terminal 7n, and / or information providing systems other than the above via the wide area network N.

[0029] FIG. 4 is an exemplary diagram of the functional block configuration of the satellite positioning enhancement device 1 constructed by the CPU 101 loading one or more computer programs. In this example, the satellite positioning enhancement device 1 operates as a data acquisition unit 201, an ionospheric penetration point identification unit 202, an ionospheric delay amount calculation unit 203, an electron density distribution model generation unit 204, a total electron content prediction model construction unit 205, and a parameter generation unit 206. The ionospheric penetration point identification unit 202, the ionospheric delay amount calculation unit 203, and the electron density distribution model generation unit 204 function as "electron density distribution model generation means". Further, the total electron content prediction model construction unit 205 functions as "prediction means". The parameter generation unit 206 functions as "correction information generation means".

[0030] The data acquisition unit 201 is referred to when generating the new correction parameter ECP and acquires various data to be processed via an input / output device (not shown) or the communication device 11. One of the data to be acquired is the observation data and the navigation message obtained from the ranging signals (e.g., L1 wave and L2 wave) from a plurality of positioning satellites 2n observed at a plurality of GNSS observation stations. The data acquisition unit 201 acquires such observation data and navigation messages from MIRAI4 or each observation station 5n and stores them in the storage 104 in association with the observation station identification information, the reception point coordinates, and the reception date and time, respectively.

[0031] Based on each navigation message acquired by the data acquisition unit 201, the ionospheric piercing point identification unit 202 calculates the position of the positioning satellite 2n existing above the observation time of the observation data, and identifies the ionospheric piercing point (pierce point) based on the calculated position of the positioning satellite 2n and the reception point coordinates.

[0032] The ionospheric delay amount calculation unit 203 calculates the ionospheric delay amount of the satellite signal passing through the ionospheric piercing point. Specifically, using the observation data (pseudo-range and carrier phase) obtained from the ranging signals of a plurality of frequencies (e.g., L1 wave and L2 wave), the ionospheric delay amount on the radio wave propagation path connecting the positioning satellite 2n and the observation station 5n is calculated. A known method can be used for the calculation of the ionospheric delay amount.

[0033] The electron density distribution model generation unit 204 calculates the total electron number in the line-of-sight direction at the ionospheric piercing point identified by the ionospheric piercing point identification unit 203 by adding and subtracting the satellite frequency bias and the receiver frequency bias, i.e., the circuit delay time difference occurring between satellite signals of different frequencies, to the calculation result of the ionospheric delay amount calculation unit 203.

[0034] The electron density distribution model generation unit 204 also calculates the vertical total electron number, which is the total electron number in the ionosphere in the zenith direction at the ionospheric penetration point, from the positional relationship between the receiving point coordinates of the observation station 5n and the positioning satellite 2n. Specifically, it calculates the angle formed by the antenna installed at the observation station and the ionospheric penetration point, and converts the total electron number in the line-of-sight direction to the "vertical total electron number" based on this angle. Then, using the calculated vertical total electron number, it generates an electron density distribution model at the first time point expanded to a wide-area ionosphere wider than the receiving point coordinates of the observation station using spherical harmonic functions. The degree of the wide area can be determined according to the positions of the positioning terminal 3n and the information terminal 7n operated by the user. The spherical harmonic function is a known function expressed in a spherical coordinate system of the distance between the receiving point of the observation station 5n and the positioning satellite 2n, the polar angle of the earth, and the azimuth angle based on the receiving point.

[0035] Note that when correcting the ionospheric error, for applications in time periods without sunlight incidence such as at night, the conversion from the total electron number in the line-of-sight direction to the vertical total electron number can be omitted.

[0036] The wide-area ionosphere includes the ionosphere existing over the ocean away from the observation station 5n that has acquired observation data and navigation messages. Alternatively, the wide-area ionosphere may be the ionosphere of the entire earth. Also, depending on the range (area) to be widened, a known spherical cap harmonic function may be used instead of the spherical harmonic function.

[0037] The total electron number prediction model construction unit 205 constructs a total electron number prediction model at a predetermined time in the future (for example, any time from 1 hour to 24 hours) from the observation time using the electron density distribution model of the wide-area ionosphere generated by the electron density distribution model generation unit 204.

[0038] For constructing the total electron number prediction model, various machine learning models can be used. In this example, an example of using a multi-convolution long short-term memory (Multi-Convolution LSTM) is shown. The multi-convolution LSTM is a model that can simultaneously learn various spatial features and temporal features in the analysis of multi-dimensional time series data and sequence data, and has the advantage of being able to improve the accuracy of future prediction of complex data patterns such as the electron density distribution in the constantly changing ionosphere. The multi-convolution LSTM is a neural network suitable for learning data with long-term dependencies. In other words, it can be said to be an excellent model that suppresses the increase in processing load by continuously retaining important information and forgetting unnecessary information. A schematic diagram of the processing in the total electron number prediction model construction unit 205 centered on such a multi-convolution LSTM is shown in FIG. 5.

[0039] Referring to FIG. 5, the total electron number prediction model construction unit 205 includes, in addition to the vertical total electron number of the ionosphere converted by the electron density distribution model generation unit 204, for example, time series data of atmospheric environment features such as solar activity and geomagnetic activity from the GFZ6 website, and data of solar altitude angles or data associated therewith as input data. Then, the variation amount 52 at regular intervals is calculated from the vertical total electron number 51, and the solar altitude coefficient 54 calculated from the solar altitude angle is multiplied by the obtained time series data 53 of the atmospheric environment features to calculate three-dimensional data (feature quantities considering date and time, latitude, and longitude) 55 expanded on the spherical surface. The solar altitude coefficient 54 is a coefficient that, for example, sets the value to "1" when the solar altitude angle is 90 degrees and "0" when the solar altitude angle is 0 degrees, and is stored in advance in the storage 104 as one of the conversion tables.

[0040] Developing the time-series data of the atmospheric environmental characteristic quantities as three-dimensional data 55 is an improvement measure for the fact that the conventional model does not consider the direction of the sun, etc. at the receiving point coordinates at the observation time. Since extreme ultraviolet rays radiated from the sun affect only the daytime side of the earth, the total electron number prediction model construction unit 205 sets the coefficient to "0", for example, for the nighttime side, and does not use parameters such as atmospheric environmental characteristic quantities and solar altitude angles in the calculation of the vertical total electron number prediction in subsequent operations.

[0041] The multi-convolution LSTM performs convolution processing on these data using the convolution filter 56. At this time, by using convolution filters 56 of a plurality of sizes, small, medium, and large, spatial features 57 of the ionosphere on a small scale, medium scale, and large scale are extracted respectively. This is an improvement measure for the fact that the above-mentioned conventional model could not sufficiently predict the increase, decrease, and displacement conditions of the total electron number in the horizontal direction and the line-of-sight direction. As a result, it is possible to respond to the prediction of the increase and decrease of the electron number and the displacement condition of the electron density distribution in the ionosphere at various spatial scales over sparsely populated areas and over the ocean where observation stations are not installed.

[0042] The multi-convolution LSTM then performs convolution processing on the small, medium, and large-scale spatial features of the ionosphere using the convolution filter 58, and integrates the feature quantities of each scale into one, thereby outputting the prediction result of the fluctuation amount 60 of the total electron number in the future by a predetermined time from the already calculated prediction result.

[0043] In addition, in FIG. 5, for the sake of convenience of explanation, one convolution filter 56 each of small, medium, and large is shown, but by preparing more convolution filters and generating the spatial features 57 of each scale by the number of convolution filters, a model corresponding to more diverse situations can be obtained.

[0044] The total electron number prediction model construction unit 205 then outputs a predicted value 61 of the vertical total electron number in the future at a predetermined time by adding the variation amount 60 to the most recent vertical total electron number or electron density distribution 59. As an example, FIG. 6 shows the results of comparing the predicted value 61 of the vertical total electron number at a point at 5°N and 100°E with the actual value (true value) over time.

[0045] The upper part of FIG. 6 is a graph comparing the predicted value (Predicted(1hour)) 61 one hour ahead of the vertical total electron number (VTEC) with the true value (Realtime Data) 62 on the vertical axis, and the horizontal axis indicates the date and time. The lower part of FIG. 6 is the absolute value of the difference (VTEC Difference) obtained by subtracting the true value 62 from the predicted value 61. From FIG. 6, it can be seen that according to this embodiment, the predicted value 61 one hour later is infinitely close to the true value 62 at any time zone, and the error is small.

[0046] By using the multi - convolutional LSTM in this way, it is possible to suppress the decrease in prediction accuracy due to the time lag from a certain point in time (the first point in time) to a point in time in the future at a predetermined time (the second point). Note that this embodiment is not intended to be limited to the example using the multi - convolutional LSTM, and other machine learning models that can be expected to have the same operational effects can be used.

[0047] The parameter generation unit 206 generates a new correction parameter ECP optimized for each region by fitting the total electron number prediction model constructed by the total electron number prediction model construction unit 205 to the Klobuchar model. For example, the new correction parameter ECP for the area A1 near Japan is generated using the data within "Japan area" in FIG. 2. Similarly, the new correction parameter ECP for the Asia - Oceania area A2 is generated using the data within "Wide area" in FIG. 2. Thereby, it is possible to quickly generate a more refined new correction parameter ECP than using all the data including other regions. The user who operates the positioning terminal 3n and the information terminal 7n will use different regions according to the positioned location. For example, when in the Tsugaru Strait, the user may use the new correction parameter ECP for the vicinity of Japan area A1, and when on the Australia route, the user may use the new correction parameter ECP for the Asia - Oceania area A2.

[0048] As a model for optimization for each region, the well - known Levenberg - Marquardt method can be used. The Levenberg - Marquardt method is a method of an algorithm for solving non - linear least - squares problems. For example, the parameters (α 0~3 , β 0~3 ) that minimize the difference between the total vertical electron number calculated using the existing correction parameters and the above - mentioned predicted value 61 can be generated. Alternatively, the parameters (α ) that minimize the difference between the total vertical electron number calculated using the existing correction parameters and the above - mentioned predicted value 61 can be generated. Alternatively, the parameters (α 0~3 , β 0~3 ) can be generated so that the sum of the squares of the differences between the ionospheric delay obtained by actual observation and the ionospheric delay calculated by the Klobuchar model is minimized.

[0049] Note that the description of the algorithm (calculation using the Jacobian matrix) itself for generating the correction parameters (α 0~3 , β 0~3 ) of the Klobuchar model using the Levenberg - Marquardt method is described in detail in Non - Patent Documents 1 and 3 (page 146), so such descriptions can be referred to.

[0050] The generated new correction parameter ECP is loaded onto a satellite signal with the same frequency as GPS or QZSS via the communication device 11 and distributed to the positioning terminal 5n via the positioning satellite 2n existing over the specified region. Thereby, the user can perform high - precision single - code positioning using the existing positioning terminal 5n and information terminal 7n.

[0051] [Correction Information Generation Method] Next, an example of a correction information generation method executed by the satellite positioning augmentation device 1 configured as described above will be described with reference to FIG. 7. As described with reference to FIGS. 1 and 3, the satellite positioning augmentation device 1 is a computer controlled by a computer program.

[0052] Referring to FIG. 7, the satellite positioning augmentation device 1 (data acquisition unit 201) controlled by a computer program acquires one or more, in this example, two received signals (for example, L1 signal, L2 signal) from the positioning satellites 2n observed at any GNSS observation station in the first area (for example, the area A1 around Japan in FIG. 2) at the first time point (any date and time) (S1). Here, the observation data and navigation messages from a plurality of positioning satellites 2n observed at the GNSS observation station may be acquired through MIRAI4. The GNSS observation station receives satellite signals of multiple frequencies, and distributes the observation data, navigation messages, and reception point coordinates received at the observation station in real time via a wide area network. The data acquisition unit 201 acquires such observation data, navigation messages, and reception point coordinates from MIRAI4 or the observation station 5n, and stores them in association with the reception point coordinates and reception date and time, respectively. The reception point coordinates of the GNSS observation station may use the coordinates stored in the data file of the storage 104.

[0053] The satellite positioning augmentation device 1 (ionospheric piercing point identification unit 202) calculates the positions of the positioning satellites 2n existing above the observation station at the time of observation based on the navigation messages from the GNSS observation station, and calculates the positions of the calculated positioning satellites 2n and the reception point coordinates (known) (S2). Further, an ionospheric piercing point (pierce point) on the propagation path connecting the positioning satellite 2n and the reception point at the first time point is identified (S3). S2 and S3 can be realized using known methods in this technical field.

[0054] The satellite positioning augmentation device 1 (ionospheric delay calculation unit 203) calculates the ionospheric delay amount on the radio wave propagation path at the first time point by connecting the positions of the observation station and the positioning satellites 2n using the above observation data at the first time point, and transfers the process to the electron density distribution model generation unit 204. The satellite positioning augmentation device 1 (electron density distribution model generation unit 204) converts the ionospheric delay amount into the total electron number on the radio wave propagation path, and generates an electron density distribution model of the wide-area ionosphere using the spherical harmonic function or the spherical cap harmonic function with this total electron number (S4).

[0055] The satellite positioning augmentation device 1 (total electron number prediction model construction unit 205) constructs a total electron number prediction model by machine learning, in this example, a multi-convolutional LSTM model, and inputs time-series data of atmospheric environment characteristics and data of solar altitude angles, etc. into the multi-convolutional LSTM, thereby deriving a predicted value of the total electron number at the second time point in the future for a predetermined time (for example, 1 hour, 2 hours... 24 hours) (S5).

[0056] The satellite positioning augmentation device 1 (parameter generation unit 206) applies the above predicted value to the Klobuchar model, and generates correction information optimized for the specified areas (the Japan area A1 and the Asia-Oceania area A2 around Japan in FIG. 2), in this example, new correction parameters for each specified area (S6).

[0057] FIG. 8 is a diagram showing the single-code positioning error when using the observation data (at 30-second intervals for 24 hours) of the Miyakojima observation station of QZSS acquired from MIRAI4. In the figure, (a) is the case of not using the ionospheric delay parameter, (b) is the case of using the existing correction parameter transmitted by GPS, (c) is the case of using the existing correction parameter transmitted by QZSS, and (d) is an example of using the new correction parameter. It shows that the closer to the center of the grid points in the figure, the closer to the true value. As is clear from these figures, it can be seen that the distribution of the positioning results is more concentrated around the true position and the correction accuracy is higher in the case of using the new correction parameter in (d) than in the cases of (a) to (c).

[0058] Thus, in the satellite positioning enhancement device 1 of the present embodiment, since an electron density distribution broadened using spherical harmonic functions or the like is generated, even in positioning in an area with few observation stations (reception points), the ionospheric error can be appropriately corrected.

[0059] In the present embodiment, also, time-series data of atmospheric environment characteristics, data of solar altitude angles, etc. are input into a multi-convolutional LSTM to derive predicted values of the total electron count in the future for a predetermined time (for example, 1 hour, 2 hours... 24 hours). Therefore, the influence of time lag is mitigated, and a new correction parameter ECP for enabling high-precision positioning can be provided to satellite positioning users at one frequency.

[0060] In the present embodiment, also, observation data can be selectively used according to the region, and a new correction parameter ECP optimized for each region can be generated. Therefore, it is possible to flexibly respond to cost adjustment of information processing and positioning demand.

[0061] Note that, in the present embodiment, the satellite positioning enhancement device 1 has been described in a mode realized by one computer. However, the satellite positioning enhancement device 1 may be a mode in which a computer program is shared by a plurality of computers and the functional blocks shown in FIG. 4 are realized in a distributed or collaborative manner.

Explanation of Reference Numerals

[0062] 1... Satellite positioning support device 2n... Positioning satellite ECP... New correction parameter 3n... Positioning terminal 5n... Observation station 7n... Information terminal 11... Communication device 204... Electron density distribution model generation unit 205... Total electron count prediction model construction unit 206... Parameter generation unit

Claims

1. an electron density distribution model generating means for calculating a total number of electrons on a radio wave propagation path connecting one or more positioning satellites and an observation station based on a received signal from the positioning satellite and observed at the observation station, and generating an electron density distribution model of the wide-area ionosphere by a spherical harmonic function or a spherical crown harmonic function using the total number of electrons; a prediction means for inputting atmospheric environmental characteristics including the solar altitude angle at the time of observation and the electron density distribution model into a total electron number prediction model constructed by machine learning, and deriving a prediction value of the total electron number in the wide-area ionosphere for a predetermined time period into the future; a correction information generating means for generating correction information for correcting an ionospheric error for a predetermined time in the future using the predicted value.

2. The atmospheric environment characteristics further include time series data of at least one of solar activity and geomagnetic activity.

2. The computer program product of claim 1.

3. the electron density distribution model generating means calculates a total number of electrons in the line of sight direction at the ionospheric penetration point by adding and subtracting a circuit delay time difference occurring between signals of different frequencies, and calculates, as the total number of electrons, a total number of vertical electrons, which is the total number of electrons in the ionosphere in the zenith direction at the ionospheric penetration point, from a positional relationship between the receiving point coordinates of the observation station and the positioning satellite; 2. The computer program product of claim 1.

4. The wide area ionosphere is the ionosphere existing above a specified region.

2. The computer program product of claim 1.

5. The wide area ionosphere is the ionosphere of the entire Earth.

2. The computer program product of claim 1.

6. the total electron number prediction model is a multi-convolution LSTM model that extracts features of local and global spatial variations and features of time series variations in the ionosphere by performing a convolution process on the electron density distribution model using convolution filters of different sizes, and combines the extracted features to output the amount of variation in the electron density distribution model after the predetermined time has elapsed; 2. The computer program product of claim 1.

7. the prediction means derives the predicted value by adding the amount of variation to the electron density distribution model before a predetermined time has elapsed.

7. A computer program according to claim 6.

8. the correction information generation means generates the correction information optimized for each predetermined region by applying the predicted value to a Klobuchar model. A computer program according to any one of claims 1 to 7.

9. an electron density distribution model generating means for calculating a total number of electrons on a radio wave propagation path connecting a positioning satellite and an observation station based on one or more received signals from the positioning satellite observed at the observation station, and generating an electron density distribution model of the wide-area ionosphere by a spherical harmonic function or a spherical crown harmonic function using the total number of electrons; a prediction means for inputting atmospheric environment characteristics including a solar altitude angle at the time of observation and the electron density distribution model into a total electron number prediction model constructed by machine learning, and deriving a prediction value of the total electron number in the wide-area ionosphere for a predetermined time in the future; a correction information generating means for generating correction information for correcting the ionospheric error for the predetermined time in the future using the predicted value; A satellite positioning augmentation device comprising:

10. 1. A computer-implemented method controlled by a computer program, comprising: acquiring one or more received signals from a positioning satellite observed at an observation station; calculating a total number of electrons on a radio wave propagation path connecting the positioning satellite and the observation station based on the acquired received signal, and generating an electron density distribution model of the wide-area ionosphere by a spherical harmonic function or a spherical crown harmonic function using the calculated total number of electrons; inputting atmospheric environment characteristics including a solar altitude angle at the time of observation and the electron density distribution model into a total electron number prediction model constructed by machine learning, and deriving a predicted value of the total electron number in the wide-area ionosphere for a predetermined time into the future; The present invention is characterized in that correction information for correcting an ionospheric error for a predetermined time in the future is generated using the predicted value. Correction information generation method.

Images