Deep learning based high-precision positioning and orientation method and system for single-beidou satellite signal receiver

Abstract

The application discloses a kind of high-precision positioning and orientation method and system of single Beidou satellite signal receiver based on deep learning, the method includes: step S1: obtaining reference station original observation value and mobile station original observation value data;Step S2: the data obtained is preprocessed;Step S3: construct mixed frequency single difference model, and utilize the model to carry out ambiguity fixing, complete baseline solution;Step S4: realize high-precision positioning and high-precision orientation.The application is processed by satellite original observation data high-quality based on deep learning and mixed frequency difference ambiguity fixing technology, eliminates the influence of various interference factors in complex urban environment and tree dense weak signal environment, realizes high-precision positioning and orientation, second-level convergence speed, and has strong anti-interference ability.

Description

Technical Field

[0001] The present invention relates to the field of satellite signal data processing technology, and in particular to a high-precision positioning and orientation method and system for a single Beidou satellite signal receiver based on deep learning. The method is specifically applied to fields such as single Beidou satellite positioning and navigation, and is particularly suitable for use in weak signal environments such as complex urban environments and densely populated trees. Background Art

[0002] Currently, single Beidou receivers generally adopt a dual-antenna design, prioritizing the Beidou B1 and B2 dual-frequency positioning and heading mode for positioning and heading. This mode offers advantages such as high speed, high ambiguity resolution, and high accuracy. However, this technology is only suitable for ideal conditions where satellite navigation signal reception is good and both the Beidou-3 B1 and B2 dual-frequency signals are available. It is not suitable for complex environments such as urban areas.

[0003] In weak signal environments such as complex urban environments and densely populated trees, the following major interference factors affect positioning accuracy and positioning time:

[0004] 1. The frequency of satellite observation data obtained by the receiver is very complex. There are differences in the tracking and capture of different frequencies of the same satellite, and some frequency tracking is lost. In the same epoch, all satellites may have single-frequency, dual-frequency, or triple-frequency data.

[0005] 2. In weak signal environments such as complex urban environments and densely populated forests, the reduction in the number of visible satellites, multipath effects, and reduced satellite observation data quality all negatively impact on-chip dynamic real-time RTK positioning accuracy. In particular, the hybrid BeiDou satellite constellation (GEO / IGSO / MEO) significantly impacts the geometric distribution of visible satellites. Achieving stable RTK positioning in a single BeiDou system is more complex than in multiple GNSS systems. Summary of the Invention

[0006] In light of this, one of the objectives of this invention is to provide a high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning. By using high-quality processing of raw satellite observation data and mixing differential ambiguity fixation technology based on deep learning, the method eliminates the influence of various interference factors in weak signal environments such as complex urban environments and densely populated areas. This method achieves high-precision positioning and orientation, converges in seconds, and exhibits strong anti-interference capabilities.

[0007] One of the purposes of the present invention is achieved through the following technical solutions:.

[0008] This high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning includes the following steps:

[0009] Step S1: Obtaining the original observation data of the base station and the original observation data of the mobile station;

[0010] Step S2: preprocessing the acquired data;

[0011] Step S3: Construct a mixed frequency single-difference model and use it to fix ambiguities and complete baseline resolution.

[0012] Step S4: Achieve high-precision positioning and high-precision orientation.

[0013] Furthermore, step S2 includes preprocessing the data using a quality control technology based on deep learning, wherein the quality control technology based on deep learning includes cycle slip detection based on a recurrent neural network (RNN) and chi-square detection based on a convolutional neural network (CNN);

[0014] The RNN-based cycle slip detection includes:

[0015] Extract residuals, signal strength, and Doppler frequency changes from the raw data as input to the recurrent neural network (RNN). Use labeled data to train the model, adjust model parameters to minimize prediction error, and use cross-validation techniques to improve the model's generalization ability. Use the trained model to predict new observations and identify cycle slip events. For identified cycle slip events, take correction, elimination, or downgrade measures to ensure the accuracy of the positioning results.

[0016] The CNN-based chi-square test includes:

[0017] The convolutional neural network (CNN) is used to train the model using labeled data. The training data contains data that conforms to the expected distribution and data that does not conform to the expected distribution, so as to effectively distinguish the two types of data. The data that does not conform to the chi-square distribution is eliminated and downgraded.

[0018] Furthermore, after preprocessing the data using deep learning-based quality control technology, it also includes: constructing single-difference observation equations and double-difference observation equations, and then performing extended Kalman filtering. After the filtering is completed, the wide-lane ambiguity is fixed. If the fixation is successful, the narrow-lane ambiguity is fixed. If not, it enters the next epoch.

[0019] Furthermore, the construction and solution method of the mixing single-difference model is as follows:

[0020] Step S31: constructing a mixed-frequency single-difference model using data from single-frequency satellites, co-frequency satellites, and multi-frequency satellites;

[0021] Step S32: Obtaining a sequential filtering floating-point solution and floating-point ambiguity;

[0022] Step S33: Perform a posteriori quality control and remove observations if they fail the check;

[0023] Step S34: If the verification is passed, the ambiguity of the entire set and the subset is fixed.

[0024] Furthermore, in step S31, the data include carrier phase observation values, pseudorange observation values ​​and broadcast ephemeris data.

[0025] Furthermore, in step S32, the sequential filtering is to filter the single-difference model using a sequential filtering method such as Kalman filtering to obtain a floating-point solution and floating-point ambiguity.

[0026] Further, in step S34,

[0027] The full set ambiguity fixation is to fix the single difference ambiguity of all satellites, and use the algorithm to fix the floating point ambiguity to an integer;

[0028] Subset ambiguity fixing is to first fix the ambiguity of some satellites and then gradually expand it to full set ambiguity fixing.

[0029] Furthermore, when further processing the result data output by the mixed-difference model, a mixed-difference ambiguity fixing method combining partial ambiguity and multiple combinations of wide and narrow lanes is used to comprehensively improve the fixing effect. The specific steps of the fixing method are as follows:

[0030] Determine whether it is multi-frequency fixed or single-frequency fixed. If it is multi-frequency fixed, use the following method 1; if it is single-frequency fixed, use the following method 2;

[0031] Method 1: First, wide lane ambiguity fixation is performed. If the fixation is successful, the corresponding result data is output to perform partial wide lane ambiguity fixation. If it fails, the process ends. If the partial wide lane ambiguity fixation is successfully completed, the corresponding result data is output to perform narrow lane ambiguity fixation. If it fails, the process ends. If the narrow lane ambiguity fixation is successfully completed, the corresponding result data is output to perform partial narrow lane ambiguity fixation. If it fails, the process ends. If the partial narrow lane ambiguity fixation is successfully completed, the ambiguity is fixed and the positioning directional gain is achieved. If not, the process ends.

[0032] Method 2: First, perform non-combined ambiguity fixation. If the fixation is successful, output the corresponding result data to perform partial non-combined ambiguity fixation. If it fails, end. If the partial non-combined ambiguity fixation is successfully completed, the fixed ambiguity and positioning directional gain are achieved. If not, end.

[0033] The second object of the present invention is to provide a high-precision positioning and orientation system for a single Beidou satellite signal receiver based on deep learning, the system comprising an ARM subsystem, a radio frequency circuit unit, a power supply unit, an interface unit, a clock unit, a WIFI unit and a Beidou satellite signal receiving antenna;

[0034] The Beidou satellite signal received by the antenna first passes through a low-noise amplifier and a power divider, and then enters the RF circuit unit through a bandpass filter. The RF circuit unit digitizes the signal and sends it to the ARM small system. By executing the computer program deployed on the calculation module in the ARM small system, the high-precision positioning and orientation method of a single Beidou satellite signal receiver based on deep learning as described above is implemented, completing real-time high-precision positioning and orientation solution.

[0035] The beneficial effects of the present invention are:

[0036] (1) Efficient signal processing algorithm: The present invention adopts a deep learning-based mixing single-difference model solution algorithm in a single Beidou receiver. Compared with traditional multi-satellite system receivers, it can more effectively achieve centimeter-level positioning accuracy, sub-degree-level orientation accuracy, second-level convergence speed and strong anti-interference capability;

[0037] (2) Strong anti-interference capability: For the reception and processing of Beidou signals, the present invention adopts a mixing single-difference model solution algorithm based on deep learning, which can maintain good stability and accuracy in weak signal environments such as complex urban environments and densely populated trees, and achieve high-precision and fast positioning and orientation;

[0038] (3) High integration, low power consumption, and obvious cost advantages: This processing board adopts a highly integrated design, based on a single-chip ARM CPU small system + necessary peripheral circuits to implement receiver design, reducing dependence on external components, reducing system complexity, and achieving single-board low power consumption, small size, and low cost capabilities;

[0039] (4) Easy to develop and integrate: The present invention is well compatible and docked with the existing BeiDou application system, and supports a variety of development environments and application platforms. It supports USB, serial port, and network port connections, allowing developers to more conveniently upgrade products and expand functions.

[0040] (5) Wide range of applications: The present invention is applicable to a variety of single Beidou application scenarios, including but not limited to single Beidou receivers, unmanned driving assistance equipment, personal navigation devices, and Internet of Things devices.

[0041] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description and the preceding claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0043] Figure 1 Schematic diagram of the method flow of the present invention;

[0044] Figure 2 This is a block diagram of the method principle of the present invention;

[0045] Figure 3 This is a block diagram of the quality control technology based on deep learning;

[0046] Figure 4 This is the principle block diagram of the construction and solution method of the mixed-frequency single-difference model;

[0047] Figure 5 This is a quantitative comparison table of the effects of using the mixed single-difference model;

[0048] Figure 6 Fixed flow chart for mixed-frequency differential ambiguity;

[0049] Figure 7 This is a schematic diagram of the overall architecture of the system;

[0050] Figure 8 This is a schematic diagram of the ARM small system architecture;

[0051] Figure 9 Implement the block diagram for the RF circuit module;

[0052] Figure 10 Schematic diagram of the power system architecture;

[0053] Figure 11 This is a schematic diagram of the architecture of the WIFI unit;

[0054] Figure 12 Schematic diagram of the clock unit architecture;

[0055] Figure 13 This is a circuit diagram of the front panel interface unit of the interface unit;

[0056] Figure 14 This is the rear panel interface unit circuit diagram of the interface unit. DETAILED DESCRIPTION

[0057] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the preferred embodiments are only for illustrating the present invention, and are not intended to limit the scope of protection of the present invention.

[0058] like Figure 1 and Figure 2 As shown, the high-precision positioning and orientation method of a single Beidou satellite signal receiver based on deep learning of the present invention includes the following steps:

[0059] Step S1: Obtaining the original observation data of the base station and the original observation data of the mobile station;

[0060] Step S2: preprocessing the acquired data;

[0061] Step S3: Construct a mixed frequency single-difference model and use it to fix ambiguities and complete baseline resolution.

[0062] Step S4: Achieve high-precision positioning and high-precision orientation.

[0063] The above steps are further described below. First, in step S2, the data is preprocessed using a quality control technology based on deep learning. The quality control technology based on deep learning includes cycle slip detection based on a recurrent neural network (RNN) and chi-square detection based on a convolutional neural network (CNN).

[0064] RNN-based cycle slip detection includes:

[0065] Considering the consistency between carrier phase and pseudorange observations, a cycle slip can be determined when the two are inconsistent. Therefore, cycle slip detection first extracts residuals, signal strength, and Doppler frequency changes from the original data as input to the recurrent neural network (RNN). The model is trained using labeled data, and model parameters are adjusted to minimize prediction errors. Cross-validation and other techniques are used to improve the model's generalization ability. The trained model is used to predict new observation data and identify cycle slip events. For identified cycle slip events, correction, elimination, or weighting are taken to ensure the accuracy of the positioning results.

[0066] CNN-based chi-square tests include:

[0067] Traditional chi-square tests, which compare observed values ​​with expected values, have low accuracy and cannot effectively identify abnormal observations. This paper leverages the powerful nonlinear fitting capabilities of convolutional neural networks (CNNs) to construct a more complex statistical test method. This method uses labeled data to train the model. The training data includes both data that conforms to the expected distribution and data that does not, effectively distinguishing between the two types of data. Data that does not conform to the chi-square distribution is eliminated and downgraded.

[0068] In weak signal environments such as complex urban environments and densely populated trees, the reduction in the total number of visible satellites, multipath effects, and the decline in the quality of satellite observation data will have a negative impact on the on-chip dynamic real-time RTK positioning accuracy; in particular, the obvious impact of the Beidou satellite hybrid constellation (GEO / IGSO / MEO) on the geometric distribution relationship of visible satellites makes it more complicated to achieve stable RTK positioning technology in a single Beidou system relative to a GNSS multi-system. The present invention addresses the problem that conventional data preprocessing technology cannot effectively and timely achieve cycle slip detection and chi-square test, and creatively achieves high-quality processing of satellite raw observation data based on deep learning, greatly improving the effectiveness of cycle slip detection and chi-square test, and achieving on-chip single Beidou RTK real-time centimeter-level positioning in weak signal environments such as complex urban environments and densely populated trees.

[0069] like Figure 3 As shown, after the data is preprocessed using deep learning-based quality control technology, it also includes: constructing single-difference observation equations and double-difference observation equations, and then performing extended Kalman filtering. After the filtering is completed, the wide-lane ambiguity is fixed. If the fixation is successful, the narrow-lane ambiguity is fixed. If not, enter the next epoch.

[0070] like Figure 4 As shown, in step S3, the construction and solution method of the mixing single-difference model is as follows:

[0071] Step S31: constructing a mixed-frequency single-difference model using data from single-frequency satellites, co-frequency satellites, and multi-frequency satellites;

[0072] Step S32: Obtaining a sequential filtering floating-point solution and floating-point ambiguity;

[0073] Step S33: Perform a posteriori quality control and remove observations if they fail the check;

[0074] Step S34: If the verification is passed, the ambiguity of the entire set and the subset is fixed.

[0075] In step S31, a mixed-frequency single-difference model is constructed using data from single-frequency satellites, co-frequency satellites, and multi-frequency satellites. The data source includes constructing a mixed-frequency single-difference model using data from single-frequency satellites, co-frequency satellites, and multi-frequency satellites. This data includes carrier phase observations, pseudorange observations, and broadcast ephemeris. For multi-frequency satellites, using carrier phase observations at different frequencies can better eliminate errors such as ionospheric delay. Error processing is performed so that the single-difference ionospheric delay and tropospheric delay can be approximated to zero when the user is close to the base station. For longer baselines, a dual-frequency ionosphere-free combination can be used to eliminate the impact of ionospheric delay.

[0076] In step S32, sequential filtering uses a sequential filtering method, such as Kalman filtering, to filter the single-difference model to obtain a floating-point solution and floating-point ambiguities. Furthermore, smoothing can be used to smooth the floating-point ambiguities over multiple epochs to reduce the effects of pseudorange noise and multipath error, thereby improving the accuracy of ambiguity estimation.

[0077] In step S33, a posteriori quality control is performed, and observations that fail verification are discarded. This verification method involves performing a posteriori quality control on the floating-point ambiguities and their residuals to check for outliers or errors. Abnormal handling includes: if verification fails, the abnormal observation is discarded and filtering and ambiguity estimation are repeated. For satellite signals with frequently abnormal signals, the observation data of that satellite can be temporarily excluded to ensure solution stability.

[0078] In step S34, global ambiguity fixing is performed on all satellite single-difference ambiguities, using algorithms such as the integer least squares method (LAMBDA method) to fix floating-point ambiguities to integers. The fixed ambiguities are used to further improve positioning accuracy and support high-precision positioning applications.

[0079] Subset ambiguity fixing allows you to first fix the ambiguities of a subset of satellites and then gradually expand to full ambiguity fixation in certain situations. Subset ambiguity fixing can improve the fix success rate, especially when observation conditions are poor or the number of satellites is small.

[0080] This technology achieves centimeter-level positioning accuracy, sub-degree-level orientation accuracy, second-level convergence speed and strong anti-interference capability through efficient error suppression of the mixed-frequency single-difference model and geometric enhancement of baseline constraints. It provides a reliable solution for high-precision navigation in weak signal environments such as complex urban environments and densely populated trees. The capability improvement brought by the implementation of the technology is as follows: Figure 5 is shown in the list.

[0081] When further processing the result data output by the mixed-frequency single-difference model, the present invention addresses the problems of the current mainstream full ambiguity fixation, such as the difficulty and low fixation rate, by adopting a mixed-frequency differential ambiguity fixation method that combines partial ambiguity and a variety of wide and narrow lane combinations to further comprehensively improve the fixation effect. The mixed-frequency differential ambiguity fixation flow chart is shown in the figure. Figure 6 As shown, in this embodiment, the specific steps of the fixing method are as follows:

[0082] Determine whether it is multi-frequency fixed or single-frequency fixed. If it is multi-frequency fixed, use the following method 1; if it is single-frequency fixed, use the following method 2;

[0083] Method 1: First, wide lane ambiguity fixation is performed. If the fixation is successful, the corresponding result data is output to perform partial wide lane ambiguity fixation. If it fails, the process ends. If the partial wide lane ambiguity fixation is successfully completed, the corresponding result data is output to perform narrow lane ambiguity fixation. If it fails, the process ends. If the narrow lane ambiguity fixation is successfully completed, the corresponding result data is output to perform partial narrow lane ambiguity fixation. If it fails, the process ends. If the partial narrow lane ambiguity fixation is successfully completed, the ambiguity is fixed and the positioning directional gain is achieved. If not, the process ends.

[0084] Method 2: First, perform non-combined ambiguity fixation. If the fixation is successful, output the corresponding result data to perform partial non-combined ambiguity fixation. If it fails, end. If the partial non-combined ambiguity fixation is successfully completed, the fixed ambiguity and positioning directional gain are achieved. If not, end.

[0085] By combining partial ambiguity selection with wide- and narrow-lane hierarchical fixation, this technology improves the ambiguity fixation success rate from the traditional FAR (Far) of 50%-70% to 85%-95%, shortens convergence time by 70%-90%, achieves positioning accuracy of 2-5cm (RMS) in dynamic scenarios, and significantly enhances robustness in environments with occlusion and interference. This technology provides an efficient and reliable ambiguity resolution solution for high-precision positioning in complex urban environments and weak signal conditions such as densely populated areas.

[0086] In the algorithm design of the present invention, combined with practical problems in actual applications, targeted innovative designs are carried out in key modules such as data preprocessing, mixing single-difference model construction and ambiguity fixation in the algorithm: high-quality processing technology of satellite original observation data based on deep learning is adopted. Under the mixing single-difference algorithm model, a mixing differential ambiguity fixation method combining partial ambiguity and multiple combinations of wide and narrow lanes is adopted to achieve centimeter-level positioning accuracy, sub-degree directional accuracy, second-level convergence speed and strong anti-interference ability in weak signal environments such as complex urban environments and densely populated trees.

[0087] Based on the design ideas of the aforementioned positioning and orientation methods, such as Figure 7 As shown, the present invention also provides a high-precision positioning and orientation system for a single Beidou satellite signal receiver based on deep learning, the system includes an ARM small system, a radio frequency circuit unit, a power supply unit, an interface unit, a clock unit, a WIFI unit and a Beidou satellite signal receiving antenna;

[0088] The BeiDou satellite signal received by the antenna first passes through a low-noise amplifier and power splitter, then a bandpass filter and enters the RF circuit unit. The RF circuit unit digitizes the signal and sends it to the ARM mini-system. The computing module in the ARM mini-system uses the previously described deep learning-based high-precision positioning and orientation method for a single BeiDou satellite signal receiver to complete real-time high-precision positioning and orientation solutions. The following is a further explanation of each component of the system:

[0089] (1) ARM small system

[0090] In this embodiment, Figure 8 As shown in the figure, the minimum ARM peripheral system mainly includes 6 parts: ARM CPU, DDR3 memory, NANDFLASH, USB interface, debug serial port and debug network port.

[0091] a) ARM: The ARM MCIMX6G2CVM05AB chip serves as the core, along with peripheral clocks, resets, DDR3 memory, NAND Flash, NOR Flash, USB ports, debug serial ports, and network ports, providing the control platform for the entire module. The ARM chip has a built-in computing unit and implements a deep learning-based single-difference model solving algorithm.

[0092] b) DDR3 memory: It is a large-capacity external memory for ARM, enabling ARM to meet the needs of data operation, storage and data exchange;

[0093] c) NAND FLASH data storage: NAND FLASH with MMC interface is mainly used to store system startup program, system program and user program, with a design capacity of 32GB;

[0094] d) USB interface: Two USB interfaces, one connected to the rear panel for PC debugging download / storage data download, and the other connected to the baseband board for communication;

[0095] e) Debug serial port / communication serial port: It is composed of the TTL serial port provided by the ARM chip and the RS232 level conversion chip. It is mainly used for communicating with the PC serial port and downloading programs during debugging;

[0096] f) Debug / communication network port: There are two network ports in total. One is connected to the rear panel and is mainly used for communication with the PC network port and downloading programs during debugging; it is used as a communication Ethernet port when not in debugging mode. The other is connected to the baseband board for communication.

[0097] (2) RF circuit unit

[0098] The RF circuit receives two Beidou satellite signals. Each Beidou satellite signal first reaches the antenna module. The weak satellite signal received by the antenna is amplified by the low-noise amplifier and sent to the power splitter. Then, the satellite signal is split by the corresponding filters. The multiple satellite signals that are split enter the RF circuit. After down-conversion, amplification and analog-to-digital conversion inside the RF circuit, the digital signal is output. The implementation block diagram is shown below. Figure 9 shown.

[0099] (3) Power supply system

[0100] The power supply system mainly consists of four parts: overvoltage protection circuit, charging management circuit, combined circuit monitor, and power supply monitoring circuit. The functions of each part are as follows:

[0101] a) Overvoltage protection circuit: When the input power exceeds the set threshold, the power supply path is cut off;

[0102] b) Charging management circuit: manages the charging of the device’s lithium battery;

[0103] c) Combined power monitor: selects the power input path (external power supply or battery) and responds to external power on / off signals;

[0104] d) Power monitoring circuit: controls battery charging and receives shutdown commands from software.

[0105] The power system architecture of this embodiment is as follows Figure 10 shown.

[0106] (4) WIFI unit

[0107] The WIFI unit designed by the present invention supports 802.11b / g / n and AP. The SDHC2 controller and UART3 controller of the CPU are used to communicate with the WIFI chip WL1837MOD to realize the WIFI function. The chip supports 2.4G and 5.0G dual-band RF signals, integrates RF, power amplifier, clock, RF switch, filter, passive components and power management unit, and has an operating temperature of -40℃~85℃; supports WLAN processors and RF transceivers of IEEE standards 802.11a, 802.11b, 802.11g and 802.11n, supports 4-bit SDIO host interface, supports simultaneous configuration of WLink 8 through STA and AP; supports Bluetooth 4.1 and CSA2, and performs Bluetooth data transmission with the host through the UART interface; supports the coexistence of WIFI and Bluetooth; in order to meet the requirements of industrial-grade environment use, a temperature-compensated crystal oscillator is integrated inside the chip. In low-power mode, the CPU can turn off the power of WIFI. The module block diagram is as follows Figure 11 As shown, the protocol requirements are met.

[0108] (5) Clock unit

[0109] The 12MHz clock is used as the working clock of STM32F100C8 and is realized by crystal oscillator; 32.768KHz is used as the low-speed clock of ARM and WIFI chips and is also realized by crystal oscillator; 24MHz is used as the system clock of ARM chip and is realized by crystal oscillator; 50MHz is used as the reference clock of PHY chip and is provided by CPU; 26MHz is the reference clock of WIFI chip and is integrated inside the chip. Figure 12 shown.

[0110] (6) Interface unit

[0111] The receiver interface unit consists of two parts: the front panel interface and the rear panel interface.

[0112] like Figure 13 As shown in the figure, the front panel interface unit circuit mainly includes three parts: 6-way LED drive signal output, 6-way key signal input and 1-way I2C interface.

[0113] like Figure 14 As shown, the rear panel interface circuit mainly includes 4 parts: 1 USB interface, 2 RS232 interfaces, 1 RS485 interface and 1 10M / 100M Ethernet interface.

[0114] It should be appreciated that embodiments of the present invention can be implemented or practiced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The methods can be implemented in a computer program using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner, according to the methods and figures described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, the program can be run on a programmed application-specific integrated circuit for this purpose.

[0115] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by the context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that can be executed by one or more processors.

[0116] Further, the methods can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, RAM, ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. In addition, the machine-readable code, or portions thereof, can be transmitted over a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning, characterized by: The method comprises the following steps: Step S1: Obtaining the original observation data of the base station and the original observation data of the mobile station; Step S2: preprocessing the acquired data; including preprocessing the data using a deep learning-based quality control technology, wherein the deep learning-based quality control technology includes cycle slip detection based on a recurrent neural network (RNN) and chi-square detection based on a convolutional neural network (CNN); The RNN-based cycle slip detection includes: Extract residuals, signal strength, and Doppler frequency changes from the raw data as input to the recurrent neural network (RNN). Use labeled data to train the model, adjust model parameters to minimize prediction error, and use cross-validation techniques to improve the model's generalization ability. Use the trained model to predict new observations and identify cycle slip events. For identified cycle slip events, take correction, elimination, or downgrade measures to ensure the accuracy of the positioning results. The CNN-based chi-square test includes: Using convolutional neural networks (CNNs), the model is trained using labeled data. The training data includes data that conforms to the expected distribution and data that does not conform to the expected distribution, effectively distinguishing between the two types of data. Data that does not conform to the chi-square distribution is eliminated and downgraded. Step S3: Construct a mixed-frequency single-difference model, and use the model to fix ambiguities and complete baseline resolution. The construction and resolution methods of the mixed-frequency single-difference model are as follows: Step S31: constructing a mixed-frequency single-difference model using data from single-frequency satellites, co-frequency satellites, and multi-frequency satellites; Step S32: Obtaining a sequential filtering floating-point solution and floating-point ambiguity; Step S33: Perform a posteriori quality control and remove observations if they fail the check; Step S34: If the verification is passed, fix the ambiguity of the full set and the subset; Step S4: Achieve high-precision positioning and high-precision orientation.

2. The high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning according to claim 1 is characterized in that: After preprocessing the data using deep learning-based quality control technology, it also includes: constructing single-difference observation equations and double-difference observation equations, and then performing extended Kalman filtering. After filtering is completed, the wide-lane ambiguity is fixed. If the fixation is successful, the narrow-lane ambiguity is fixed. If not, it enters the next epoch.

3. The high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning according to claim 1, characterized in that: In step S31, the data includes carrier phase observation values, pseudorange observation values ​​and broadcast ephemeris data.

4. The high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning according to claim 1, characterized in that: In the step S32, the sequential filtering is to filter the single-difference model using a Kalman filter sequential filtering method to obtain a floating-point solution and a floating-point ambiguity.

5. The high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning according to claim 1, characterized in that: In step S34, the full set ambiguity fixing is to fix the single difference ambiguity of all satellites, and the floating point ambiguity is fixed to an integer using an algorithm; the subset ambiguity fixing is to first fix the ambiguity of some satellites and then gradually expand it to the full set ambiguity fixing.

6. The high-precision positioning and orientation method for a single Beidou satellite signal receiver based on deep learning according to claim 1, characterized in that: When further processing the output data of the mixed-difference model, a mixed-difference ambiguity fixation method combining partial ambiguity and various combinations of wide and narrow lanes is used to comprehensively improve the fixation effect. The specific steps of the fixation method are as follows: Determine whether it is multi-frequency fixed or single-frequency fixed. If it is multi-frequency fixed, use the following method 1; if it is single-frequency fixed, use the following method 2; Method 1: First, wide-lane ambiguity fixation is performed. If the fixation is successful, the corresponding result data is output to perform partial wide-lane ambiguity fixation. If it fails, the process ends. If the partial wide-lane ambiguity fixation is completed successfully, the corresponding result data is output to perform narrow-lane ambiguity fixation. If it fails, the process ends. If the narrow-lane ambiguity fixation is completed successfully, the corresponding result data is output to perform partial narrow-lane ambiguity fixation. If it fails, the process ends. If the partial narrow-lane ambiguity fixation is completed successfully, the ambiguity is fixed and the positioning directional gain is achieved. If not, the process ends. Method 2: First, perform non-combined ambiguity fixation. If the fixation is successful, output the corresponding result data to perform partial non-combined ambiguity fixation. If it fails, end. If the partial non-combined ambiguity fixation is successfully completed, the fixed ambiguity and positioning directional gain are achieved. If not, end.

7. A high-precision positioning and orientation system for a single Beidou satellite signal receiver based on deep learning, characterized by: The system includes an ARM subsystem, a radio frequency circuit unit, a power supply unit, an interface unit, a clock unit, a WIFI unit and a Beidou satellite signal receiving antenna; The Beidou satellite signal received by the antenna first passes through a low-noise amplifier and a power divider, and then enters the RF circuit unit through a bandpass filter. The RF circuit unit digitizes the signal and sends it to the ARM small system. By executing the computer program deployed on the operation module in the ARM small system, the high-precision positioning and orientation method of a single Beidou satellite signal receiver based on deep learning as described in any one of claims 1 to 6 is implemented to complete real-time high-precision positioning and orientation solution.

Images