Vehicle positioning method, device and equipment based on beidou satellite signal and medium

Abstract

The application provides a vehicle positioning method and device based on Beidou satellite signals, equipment and medium. The method acquires Beidou satellite signals. If a reference station signal is received, a reference station positioning mode is entered, the Beidou satellite signals are processed according to a single-frequency real-time differential technology to acquire satellite positioning data, the satellite positioning data, inertial navigation data, mileage data and directional data are input into a tightly coupled positioning model for fusion processing to output fusion positioning data. High-frequency errors and low-frequency errors generated in the positioning process are acquired, the fusion positioning data is corrected according to the high-frequency errors and low-frequency errors to generate vehicle positioning results. The method improves accident determination efficiency, optimizes UBI insurance billing and guarantees vehicle asset safety in the financial field, shortens emergency rescue response time, enhances the reliability of old-age care and helps precise interaction of vehicle medical data in the medical and health care fields, and has the advantages of autonomous control and low power consumption expansion.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, and in particular to a vehicle positioning method and device based on Beidou satellite signals, equipment and medium. BACKGROUND

[0002] With the development of intelligent transportation, automatic driving and Internet of Things technology, the application demand of vehicle positioning system in the fields of finance, medical and health care, and the like is increasingly urgent. In the field of finance, the positioning accuracy of vehicles directly affects the determination of accident liability, the tracking of stolen vehicles, and the charging of UBI (insurance based on usage), and the traditional positioning system is prone to cause claims disputes due to poor signal stability in complex environments (such as the "valley effect" between urban high-rise buildings and tunnel shielding), resulting in positioning errors of meters. In the field of medical and health care, for the vehicle monitoring of the elderly group or special patients and the emergency rescue scene, the positioning system needs to continuously provide high-precision position and heading information in complex road conditions, and the inertial navigation error accumulates significantly when the signal is lost for a long time (such as in tunnels), which cannot meet the demand of real-time and accurate positioning for emergency rescue. The current vehicle positioning technology in the industry mainly has the following defects:

[0003] 1. Poor adaptability to complex environments: relying on GPS or Beidou / GPS dual-mode system, the positioning accuracy is reduced to meters in urban canyons, tunnels and other scenes affected by multipath effect and shielding, which cannot meet the requirements of financial insurance for high-precision restoration of accident scenes and medical rescue for real-time position tracking.

[0004] 2. Low multi-source fusion efficiency: the traditional tight combination model does not effectively integrate directional data of double antennas, resulting in heading drift under viaducts, and does not distinguish the error characteristics of high and low frequencies, so the correction efficiency is low and stable positioning results cannot be continuously output in dynamic driving.

[0005] 3. Poor service expansion of reference station: the server has high load and large response delay when a large number of concurrent services are provided, which makes it difficult to support the demand of real-time monitoring of a large number of vehicles in the insurance industry and synchronous positioning of multiple devices in the elderly care monitoring system.

[0006] Therefore, there is an urgent need for a method to solve at least one of the above problems. SUMMARY

[0007] The present application provides a vehicle positioning method, device, equipment and medium based on Beidou satellite signals, aiming to solve the defects of the current vehicle positioning technology in the industry, such as poor adaptability to complex environments, low multi-source fusion efficiency and poor service expansion of reference station.

[0008] In a first aspect, the present application provides a vehicle positioning method based on Beidou satellite signals, comprising:

[0009] The Beidou satellite signals are acquired, and multi-frequency non-coherent accumulation processing is performed on the acquired Beidou satellite signals to enhance weak signal acquisition.

[0010] Inertial navigation data collected by an inertial navigation device carried by the vehicle, mileage data collected by a wheel odometer, and directional data collected by a dual-antenna module are acquired.

[0011] If a reference station signal is received, a reference station positioning mode is entered, the Beidou satellite signals are processed according to a single-frequency real-time differential technology to acquire satellite positioning data, the satellite positioning data, the inertial navigation data, the mileage data, and the directional data are input into a preset tightly coupled positioning model for fusion processing, and fusion positioning data is output.

[0012] High-frequency errors and low-frequency errors generated in the positioning process are acquired, the fusion positioning data is corrected according to the high-frequency errors and the low-frequency errors, and a vehicle positioning result is generated according to the corrected fusion positioning data.

[0013] In some embodiments, the Beidou satellite signals include Beidou-3 new frequency point signals; the multi-frequency non-coherent accumulation processing performed on the acquired Beidou satellite signals to enhance weak signal acquisition includes: performing frequency segment processing on the Beidou-3 new frequency point signals according to corresponding Beidou-3 new frequency points, performing energy accumulation on the Beidou-3 new frequency point signals in each frequency segment, and dynamically adjusting the accumulation time according to the signal strength to acquire an accumulation result; and performing weighted merging on the accumulation results of each frequency segment to acquire enhanced Beidou satellite signals.

[0014] In some embodiments, before the inertial navigation data collected by the inertial navigation device carried by the vehicle, the mileage data collected by the wheel odometer, and the directional data collected by the dual-antenna module are acquired, the Beidou satellite signals are subjected to electromagnetic interference suppression through a preset adaptive narrowband interference suppression circuit, the frequency range of narrowband interference is identified through real-time monitoring of the signal spectrum, and a notch filter corresponding to the interference frequency is dynamically generated; the Beidou satellite signals are filtered through the notch filter to remove the narrowband interference components.

[0015] In some embodiments, before the satellite positioning data, the inertial navigation data, the mileage data, and the directional data are input into the preset tightly coupled positioning model for fusion processing, if a reference station signal is not received, a reference station positioning mode is entered, the Beidou satellite signals are processed using a single-frequency precise point positioning technology to acquire satellite positioning data, and the satellite positioning data, the inertial navigation data, the mileage data, and the directional data are input into the tightly coupled positioning model.

[0016] In some embodiments, the fusion processing of the satellite positioning data, the inertial navigation data, the mileage data and the directional data in the preset tight combination positioning model to output the fusion positioning data comprises: establishing a state space model of the satellite positioning data, the inertial navigation data, the mileage data and the directional data in the tight combination positioning model; obtaining the speed and the heading corresponding to the inertial navigation data; predicting the vehicle position, the speed and the heading as state variables and taking the position information corresponding to the satellite positioning data and the displacement information corresponding to the mileage data as observation variables, iteratively updating the state variables by a Kalman filtering algorithm, and introducing the directional data of the double-antenna to correct and constrain the heading angle, to output the fusion positioning data.

[0017] In some embodiments, the high-frequency error and the low-frequency error generated in the positioning process are obtained, comprising: classifying the errors in the positioning process according to time characteristics, classifying the receiver clock error and the signal acquisition noise as the high-frequency error, and estimating the high-frequency error cycle by cycle through the satellite signal observation values collected in real time; classifying the ionospheric delay and the tropospheric delay as the low-frequency error, and trend predicting the low-frequency error according to the historical observation data of the reference station network and the real-time meteorological data.

[0018] In some embodiments, the fusion positioning data is corrected according to the high-frequency error and the low-frequency error to generate a vehicle positioning result according to the corrected fusion positioning data, comprising: filtering and correcting the high-frequency error by a real-time Kalman filtering algorithm to compensate into the position, speed and heading parameters corresponding to the fusion positioning data; for the low-frequency error, the correction information of the regular grid points of the low-frequency error broadcast by the reference station network is received, the correction value of the adjacent grid points is matched according to the real-time position of the vehicle, the ionospheric and tropospheric delay errors in the fusion positioning data are weighted and corrected, and the vehicle positioning result containing the position coordinates and the heading angle is generated.

[0019] In a second aspect, the application provides a vehicle-mounted positioning device based on Beidou satellite signals, comprising:

[0020] A signal acquisition unit is configured to acquire Beidou satellite signals and perform multi-frequency non-coherent accumulation processing on the acquired Beidou satellite signals to enhance weak signal acquisition.

[0021] A data acquisition unit is configured to acquire inertial navigation data collected by an inertial navigation device carried by a vehicle, mileage data collected by a wheel odometer and directional data collected by a double-antenna module.

[0022] a fusion output unit, configured to: if a reference station signal is received, enter a reference station positioning mode, process the Beidou satellite signals according to a single-frequency real-time kinematic differential technique, obtain satellite positioning data, input the satellite positioning data, inertial navigation data, mileage data and directional data into a preset tightly coupled positioning model for fusion processing, and output fusion positioning data;

[0023] a result generation unit, configured to: obtain high-frequency errors and low-frequency errors generated in the positioning process, correct the fusion positioning data according to the high-frequency errors and the low-frequency errors, and generate a vehicle positioning result according to the corrected fusion positioning data.

[0024] In a third aspect, the present application also provides a computer device, comprising:

[0025] a memory and a processor;

[0026] the memory is configured to store a computer program;

[0027] the processor is configured to execute the computer program and implement the steps of the vehicle positioning method based on Beidou satellite signals according to the first aspect when executing the computer program.

[0028] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is configured to make a processor implement the steps of the vehicle positioning method based on Beidou satellite signals according to the first aspect when executed by the processor.

[0029] The vehicle positioning method based on Beidou satellite signals, the device, the equipment and the medium provided by the embodiments of the present application are aimed at obtaining signals containing new frequency points of Beidou-3, enhancing weak signal acquisition capability through multi-frequency non-coherent accumulation, suppressing electromagnetic interference through an adaptive anti-narrowband interference circuit, and improving signal quality in complex environments. Inertial navigation data, wheel mileage data and double-antenna directional data are collected, and single-frequency real-time kinematic differential (RTK) or single-frequency precise point positioning (PPP) technology is used to process Beidou signals according to whether a reference station signal is received. Multi-source data is fused through a preset tightly coupled positioning model, and a double-antenna directional constraint is introduced to solve the problem of heading drift under a viaduct. High-frequency errors (receiver clock errors) are estimated in real time, low-frequency errors (ionospheric / tropospheric delays) are weighted and predicted, and reference station network broadcast correction information is used to reduce server load through virtual grid data broadcasting technology, and finally a high-precision positioning result is generated.

[0030] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0032] Figure 1 is a step schematic flow chart of a vehicle positioning method based on Beidou satellite signals provided by an embodiment of the present application;

[0033] Figure 2 is a step schematic flow chart of a Beidou satellite signal enhancement method provided by an embodiment of the present application;

[0034] Figure 3 is a step schematic flow chart of another vehicle positioning method based on Beidou satellite signals provided by an embodiment of the present application;

[0035] Figure 4 is a structural schematic diagram of a vehicle positioning device based on Beidou satellite signals provided by an embodiment of the present application;

[0036] Figure 5 is a structural schematic block diagram of a computer device provided by an embodiment of the present application.

[0037] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the scope of protection of the present application.

[0039] The flow charts shown in the drawings are only exemplary, and do not necessarily include all the contents and operations / steps, and do not necessarily be executed in the described order. For example, some operations / steps can be decomposed, combined or partially merged, so the actual execution order can be changed according to the actual situation.

[0040] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, the same items or similar items with basically the same functions and effects are distinguished by using "first", "second" and the like. Those skilled in the art can understand that "first", "second" and the like do not limit the quantity and execution order, and "first", "second" and the like do not necessarily mean different.

[0041] It is to be understood that the terminology used herein in the specification and the appended claims is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0042] It is also to be understood that the terminology "and / or" as used herein in the specification and in the claims, is used to describe one or more of the stated features.

[0043] Some embodiments of the present application will now be described in detail in connection with the accompanying drawings. The following embodiments and features are merely exemplary and can be combined with each other in any way possible.

[0044] With the development of intelligent transportation, autonomous driving and Internet of Things technology, the application demand of vehicle positioning system in the field of finance and insurance, medical and health care, and the like is increasingly urgent. In the field of finance and insurance, the positioning accuracy of vehicles directly affects the determination of accident liability, the tracking of stolen vehicles, and the charging of UBI (insurance based on usage), and the traditional positioning system has poor signal stability in complex environments (such as the "canyon effect" between high-rise buildings in cities and tunnel shielding), resulting in positioning errors of meters, which easily leads to claims disputes; in the field of medical and health care, for the vehicle monitoring of the elderly group or special patients and emergency rescue scenarios, the positioning system needs to provide high-precision position and heading information in complex road conditions, and the inertial navigation error accumulates significantly when the signal is lost for a long time (such as in tunnels), which cannot meet the real-time and accurate positioning requirements of emergency rescue. The current vehicle positioning technology in the industry mainly has the following defects:

[0045] 1. Poor adaptability to complex environments: relying on GPS or Beidou / GPS dual-mode system, the positioning accuracy is reduced to meters in urban canyons, tunnels and other scenes affected by multipath effect and shielding, which cannot meet the requirements of finance and insurance for high-precision restoration of accident scenes and medical rescue for real-time position tracking.

[0046] 2. Low multi-source fusion efficiency: the traditional tight combination model does not effectively integrate directional data of double antennas, resulting in heading drift under viaducts, and does not distinguish the error characteristics of high and low frequencies, so the correction efficiency is low and stable positioning results cannot be continuously output in dynamic driving.

[0047] 3. Poor service expansion of reference station: the server load is high and the response delay is large when there are a large number of concurrent times, which makes it difficult to support the needs of real-time monitoring of a large number of vehicles in the insurance industry and synchronous positioning of multiple devices in the elderly care monitoring system.

[0048] To solve the above problems, the application provides a vehicle positioning method based on Beidou satellite signals, which significantly improves the positioning accuracy and reliability in complex environments through Beidou signal enhancement processing, multi-source data tight combination fusion and high-low frequency error separation correction, effectively solving the pain points in the fields of finance, insurance, medical care and old-age care.

[0049] Please refer to Figure 1 , Figure 1 is a schematic flowchart of the vehicle positioning method based on Beidou satellite signals provided by an embodiment of the application. The vehicle positioning method based on Beidou satellite signals can be implemented by a computer device, which can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, a notebook computer, a wearable device, or a robot, etc.

[0050] It should be noted that the acquisition of any information mentioned in the provided method is in accordance with relevant regulations and with the consent of the user, and does not infringe on the privacy of the user or violate relevant laws and regulations.

[0051] As Figure 1 shown, the vehicle positioning method based on Beidou satellite signals provided includes steps S101 to S104. Details are as follows:

[0052] Step S101. Acquire Beidou satellite signals, and perform multi-frequency non-coherent accumulation processing on the acquired Beidou satellite signals to enhance weak signal acquisition.

[0053] Specifically, the multi-frequency non-coherent accumulation technology is used to enhance the weak signal acquisition capability of Beidou satellites, solving the problem of insufficient signal strength in sheltered scenes such as urban canyons and tunnels. Specifically, the non-coherent energy accumulation is performed on the multi-frequency point carrier phase / pseudorange signals of the same satellite of Beidou B1I / B2I / B3I multi-frequency signals, the signal-to-noise ratio (SNR) is improved, and effective acquisition of ultra-weak signals below-160dBm is realized.

[0054] Multi-frequency signal separation and screening receives Beidou three-frequency signals (1561.098MHz, 1268.52MHz, 1207.14MHz), separates each frequency band signal through a band-pass filter, and eliminates invalid frequency bands with low signal-to-noise ratio (SNR<25dBHz).

[0055] For example, in the financial scenario, the B1I frequency band (high update rate) is preferentially reserved for real-time accident positioning, and the B2I frequency band (high accuracy) is used for post-trajectory restoration; for example, in the medical scenario, through full-band reception, the B3I frequency band (strong anti-shielding capability) signal in the tunnel is highlighted to ensure the continuity of emergency rescue signals.

[0056] The non-coherent accumulation algorithm performs energy accumulation on different frequency band signals of the same satellite, and the expression corresponding to the Beidou satellite signal includes:

[0057] ;

[0058] Wherein Ai and Bi are the amplitude of each frequency band orthogonal carrier, Sacc is the accumulated Beidou satellite signal, n is the number of frequency bands, and the number of accumulations is dynamically adjusted according to the scene (such as 8 times of accumulation in the financial scene, and 4 times of fast accumulation in the medical scene due to the need for real-time). By introducing a sliding window mechanism, the accumulation window is updated every 50 ms to avoid phase ambiguity caused by signal Doppler shift.

[0059] In the financial scene, such as at the city intersection with high accident rate, the signal capture is enhanced to accurately record the trajectory 10 seconds before the collision, solve the positioning point jump problem caused by the "valley effect", and provide continuous trajectory evidence for responsibility determination.

[0060] In the medical scene, such as when driving in a tunnel, at least 3 satellites are tracked effectively through enhanced signals to provide basic positioning input for vehicle monitoring systems and avoid rescue delays caused by signal loss.

[0061] Step S102. Obtain the inertial navigation data collected by the inertial navigation device carried by the vehicle, the mileage data collected by the wheel odometer, and the directional data collected by the dual-antenna module.

[0062] Specifically, a multi-source heterogeneous sensor fusion system is constructed by collecting inertial navigation (IMU), wheel odometer, and dual-antenna directional module data in real time. IMU provides acceleration and angular velocity (100 Hz high frequency sampling), wheel odometer calculates displacement through wheel speed pulse (resolution 0.1 meters / pulse), and dual-antenna module obtains heading angle through carrier phase difference (accuracy ±0.1°).

[0063] IMU is installed at the center of mass of the vehicle, uses MEMS devices (zero bias stability <5° / h), and the time deviation between the hardware clock synchronization module and the Beidou receiver is <1μs; the distance between the two antennas is ≥1.2 meters (baseline length), and they are respectively deployed at the front and rear ends of the roof, and the heading angle is calculated using Beidou dual-frequency carrier phase difference: θ=arctan2(Δy,Δx), wherein Δx, Δy are the baseline vectors calculated by the phase difference of the dual-antenna.

[0064] Wheel odometer pulse signals are filtered by Kalman filter to remove wheel slip noise, vehicle CAN bus real-time wheel speed correction is used in the financial scene (error <0.5%), and air pressure sensor is added to assist slope compensation in the medical scene; dual-antenna directional data is used for accident vehicle attitude judgment (such as roll angle) in the financial scene, and is used for maintaining the continuity of the heading direction (avoiding heading drift in the tunnel) in the medical scene.

[0065] In the financial scene, by focusing on calibrating dual-antenna directional data, the collision direction is judged by the heading angle mutation detection (threshold ± 15° / s) when the accident occurs, and the collision point coordinates are accurately restored in combination with the odometer displacement (error <0.5 meters).

[0066] In the medical scene, the IMU sampling rate is increased to 200Hz, real-time monitoring of vehicle acceleration and deceleration changes, combined with odometer data to construct an inertial track when the signal is lost (such as maintaining 30 seconds of effective positioning in the tunnel), providing continuous position information for emergency rescue.

[0067] Step S103. If the reference station signal is received, enter the reference station positioning mode, process the Beidou satellite signal according to the single-frequency real-time kinematic differential technology, obtain satellite positioning data, input satellite positioning data, inertial navigation data, mileage data and directional data into a preset tightly coupled positioning model for fusion processing, and output fusion positioning data.

[0068] Specifically, based on single-frequency real-time kinematic differential (RTK) technology to obtain centimeter-level satellite positioning data, through a tightly coupled model to fuse IMU, odometer, and dual-antenna data, a state equation is constructed:

[0069] ;

[0070] The state vector x includes 15-dimensional parameters such as position, velocity, attitude, and IMU error, and the measurement equation z fuses satellite pseudo-range / carrier phase, odometer displacement, and dual-antenna heading angle.

[0071] Single-frequency RTK processing receives RTCM3.3 differential data (update rate 1Hz) broadcast by the reference station, solves the floating-point solution through the single-frequency integer ambiguity fixing algorithm (L1 band), and adopts the "continuous 3-time fixed solution verification" mechanism (to avoid accidental errors) in the financial scene, and adopts the "fast ambiguity resolution" (time consumption <2 seconds) in the medical scene. The heading state is corrected by inputting the heading angle calculated by the dual-antenna as the measurement, and the IMU attitude drift is corrected (especially the heading error problem of traditional tightly coupled models under viaducts).

[0072] In the financial scene, through the model parameter adaptive adjustment triggered by increasing the accident characteristic quantity (such as the emergency braking acceleration threshold 10m / s 2 ), the Kalman filter process noise covariance is increased 5 seconds before the collision to enhance the trajectory fitting accuracy.

[0073] In the medical scene, by designing a signal loss pre-detection mechanism (triggered when the number of satellites is less than 3), the inertial / odometer dominant mode is switched to in advance to maintain the continuity of the positioning output (position error growth rate <0.1 meters / second).

[0074] In the financial scenario, in the UBI billing, a high-precision trajectory (longitude / latitude error <0.5 meters, heading error <1°) is output in real time through a tight combination model, solving the misjudgment problem of traditional positioning systems on the upper and lower layers of the viaduct (such as distinguishing between main road / auxiliary road travel mileage).

[0075] In the medical scenario, in the vehicle-mounted monitoring, the positioning accuracy of mountainous roads is enhanced by using reference station signals, and whether the vehicle deviates from the preset rescue route (such as entering non-paved roads) is judged by directional data from dual antennas.

[0076] Step S104. Obtain high-frequency errors and low-frequency errors generated in the positioning process to correct the fusion positioning data according to the high-frequency errors and low-frequency errors, and generate a vehicle positioning result according to the corrected fusion positioning data.

[0077] Specifically, the positioning errors are separated by frequency domain analysis: high-frequency errors (>1Hz, such as IMU zero bias, wheel instantaneous slip) are corrected by real-time Kalman filtering; low-frequency errors (≤1Hz, such as satellite orbit error, ionospheric delay) are extracted by sliding window Fourier transform to extract the trend item, and are compensated by polynomial fitting.

[0078] The error frequency characteristic division includes: high-frequency error sources: IMU angular rate noise (noise density determined by Allan variance analysis), odometer pulse count error (isolated pulses removed by median filtering); low-frequency error sources: satellite clock bias drift (predicted by reference station differential data), multi-path effect accumulation (consistent detection by dual-antenna phase difference).

[0079] The differentiated correction strategy includes: high-frequency correction: by designing a two-stage Kalman filter, the first stage processes the high-frequency noise of IMU and odometer (update period 10ms), and the second stage fuses satellite / dual-antenna low-frequency information (update period 100ms);

[0080] In the financial scenario, by fitting the trend item of the 30-second trajectory before and after the accident in a sliding window (5 seconds), the long-term drift is corrected (such as the cumulative error in the tunnel is reduced from 10 meters in the traditional scheme to within 1 meter);

[0081] In the medical scenario, by enabling a low-frequency error prediction model (ARIMA model trained based on historical 3-minute error data) during signal loss, the positioning trend in the tunnel (such as slope, curve-induced inertial drift compensation) is predicted.

[0082] In the financial scenario, in the accident liability determination, the low-frequency error correction preserves the trajectory splicing accuracy of different time periods (such as position jump <0.3 meters before and after crossing the tunnel), avoiding disputes in liability division due to error accumulation.

[0083] In a medical scene, during emergency rescue, high-frequency correction real-time suppresses the IMU error caused by vehicle bumping (such as the heading fluctuation is reduced from ±5° to ±1° when turning sharply), and ensures that the rescue dispatching system obtains accurate real-time position.

[0084] In some embodiments, the Beidou satellite signal includes a Beidou-3 new frequency point signal; as Figure 2 As shown, the multi-frequency non-coherent accumulation processing of the obtained Beidou satellite signal to enhance weak signal acquisition includes steps S101a to S101b.

[0085] Step S101a. The Beidou-3 new frequency point signal is processed according to the corresponding Beidou-3 new frequency point, and the energy accumulation of the Beidou-3 new frequency point signal in each frequency segment is performed. The accumulation time is dynamically adjusted according to the signal strength to obtain the accumulation result.

[0086] Step S101b. The accumulation results of each frequency segment are combined by weighting to obtain the enhanced Beidou satellite signal.

[0087] For Beidou-3 new frequency point (such as B1C, B2a, B2b) signal, the weak signal acquisition capability is improved by frequency segment energy accumulation. The core steps are: frequency point frequency segment processing → dynamic time accumulation → weighted combination, which solves the problem of signal attenuation in complex environment.

[0088] The frequency segment and dynamic accumulation are divided according to the new frequency point frequency segment: B1C (1575.42MHz±4.092MHz), B2a (1176.45MHz±14.04MHz), B2b (1207.14MHz±24MHz), and each frequency segment is independently filtered. The accumulation time is dynamically adjusted, including: when the signal strength is greater than-150dBm, the accumulation time is 50ms, when the signal strength is-150~-160dBm, the accumulation time is 100ms, and when the signal strength is less than-160dBm, the accumulation time is 200ms (the medical scene supports at least 250ms accumulation). The energy accumulation formula is: Where It, Qt are the in-phase / quadrature components of the baseband signal, and T is the number of accumulation periods.

[0089] The weighted combination strategy includes: the weight coefficient is set according to the frequency segment characteristics: B1C (high sensitivity, weight 0.4), B2a (anti-multipath, weight 0.3), B2b (strong penetration, weight 0.3); in the financial scene, the continuity of B1C frequency segment is preferentially guaranteed, which is used for high-frequency positioning (10Hz update) in the accident moment; in the medical scene, the weight of B2b frequency segment is increased to 0.5 to enhance the signal penetration capability in the tunnel (the cumulative signal acquisition rate is increased by 30%).

[0090] In the financial scene, in the urban high-rise "canyon" area, through the joint accumulation of B1C+B2a frequency band, the positioning accuracy is improved from 5 meters of traditional scheme to 1.5 meters, ensuring that the accident collision point coordinate error is less than 1 meter, and reducing the trajectory controversy in the claim dispute.

[0091] In the medical scene, in the long tunnel (>2 kilometers), through the 200ms super-long time accumulation of B2b frequency band, at least 4 satellite signals are effectively tracked, avoiding the positioning interruption of vehicle-mounted monitoring system caused by signal loss, and gaining golden time for emergency rescue.

[0092] In some embodiments, as shown in Figure 3 Before the step of acquiring the inertial navigation data collected by the inertial navigation device carried by the vehicle, the step S105 and the step S106 are further included.

[0093] Step S105. The adaptive anti-narrow-band interference circuit is used to suppress electromagnetic interference of the Beidou satellite signal. By monitoring the signal spectrum in real time, the frequency range of narrow-band interference is identified, and a notch filter corresponding to the interference frequency is dynamically generated.

[0094] Step S106. The Beidou satellite signal is filtered by the notch filter to remove the narrow-band interference component.

[0095] By real-time spectrum monitoring, narrow-band interference (such as vehicle-mounted electronic equipment and FM broadcast stray signals) is identified, and a notch filter is dynamically generated to suppress interference and improve signal purity.

[0096] The interference detection and filter generation includes a spectrum monitoring module: scanning 0-2GHz frequency band once per second, with a resolution of 100kHz, and identifying narrow-band interference frequency band through an energy threshold (> -100dBm);

[0097] Notch filter design: for the interference center frequency f0, an IIR notch filter with a bandwidth of 2MHz is generated, and the transfer function H(z)= (1-2rcosθz -1 + 2 z -2 ) / (1-2cosθz -1 + -2 z ), where θ=2πf0 / fs, r=0.95 (suppression depth > 30dB).

[0098] In the financial scenario, focus on monitoring 1560-1570MHz (GPS L1 band neighborhood interference), interference response time <50ms, to ensure that the mileage statistics during UBI billing are not interfered by vehicle-mounted Bluetooth / Wi-Fi signals; in the medical scenario, increase the monitoring of 400-500MHz (railway / industrial intercom frequency band), filter switching delay <20ms, to avoid continuous interference of intercom signals on positioning during mountain rescue.

[0099] In the financial scenario, in the dense electronic device area such as a parking lot, the signal-to-noise ratio is improved by 5dB through a notch filter, solving the position jumping problem caused by RFID reader interference in the traditional positioning system, and ensuring the coordinate accuracy (error <2m) of vehicle entry and exit records. In the medical scenario, when an ambulance passes through the MRI equipment area of a hospital, real-time suppression of 123-128MHz (MRI radio frequency interference frequency band) is performed to maintain the stability of the positioning signal and avoid navigation errors (such as missing the emergency exit) caused by interference.

[0100] In some embodiments, before the satellite positioning data, inertial navigation data, mileage data, and orientation data are input into a preset tightly coupled positioning model for fusion processing, it further includes: if no reference station signal is received, entering a reference station-free positioning mode to process Beidou satellite signals using single-frequency precise point positioning technology to obtain satellite positioning data; and inputting the satellite positioning data, inertial navigation data, mileage data, and orientation data into the tightly coupled positioning model.

[0101] When the reference station signal is lost, single-frequency precise ephemeris (IGS fast orbit, time delay 30 minutes) and clock difference products (accuracy 0.1ns) are used to solve the decimeter-level positioning result through PPP technology, which is input into the tightly coupled model.

[0102] The single-frequency PPP solving process includes: error model: considering the first-order term of ionosphere (modified by Klobuchar model), troposphere delay (Saastamoinen model), and satellite clock difference (IGS final clock difference interpolation); state vector: containing position (3D), receiver clock difference (1D), using sequential least squares estimation (update rate 1Hz);

[0103] In the financial scenario, by enabling almanac-assisted ephemeris prediction (predicting the orbit for the next 10 minutes, error <5m), the continuity of positioning in suburban accidents is ensured; in the medical scenario, by pre-storing 7-day IGS precise ephemeris, offline solving is performed when there is no network in remote mountainous areas, and the positioning accuracy is maintained within 5m (the error of the traditional scheme is >20m). The reference station signal criterion triggers the switch when no RTCM data is received for 3 consecutive periods (3 seconds), and the medical scenario increases the inertial navigation error threshold (forced switching when the speed error is >5m / s).

[0104] In the financial scenario, when the reference station signal coverage is weak in the remote section of the highway, the positioning accuracy is maintained to 3 meters by single-frequency PPP, ensuring the position continuity in the tracking of stolen vehicles, and avoiding the loss of tracking due to signal interruption.

[0105] In the medical scenario, in the mountainous area without reference station coverage, the emergency rescue positioning error is reduced from 10 meters in the traditional scheme to 5 meters by using pre-stored ephemeris and inertial navigation fusion, providing more accurate landing coordinate reference for helicopter rescue.

[0106] In some embodiments, the satellite positioning data, inertial navigation data, mileage data and directional data are input into a preset tight combination positioning model for fusion processing, and the fusion positioning data is output, including: in the tight combination positioning model, a state space model of satellite positioning data, inertial navigation data, mileage data and directional data is established; the speed and heading corresponding to the inertial navigation data are obtained; the vehicle position predicted by the inertial navigation data, the speed and heading are taken as state variables, and the position information corresponding to the satellite positioning data and the displacement information corresponding to the mileage data are taken as observation variables, the state variables are iteratively updated by Kalman filtering algorithm, and the heading angle is corrected and constrained by introducing double-antenna directional data, and the fusion positioning data is output.

[0107] By constructing a 15-dimensional state space model (position, speed, attitude, IMU error, clock error, etc.), the inertial navigation prediction state is obtained, satellite positioning and odometer are taken as observation, double-antenna directional data is introduced to correct the heading angle, and the heading drift problem under the viaduct is solved.

[0108] The corresponding state equation of the state space model includes: , wherein F is the system matrix (including the earth rotation, the IMU error model), G is the noise matrix; the observation equation: satellite pseudo-range / carrier phase residual, odometer displacement difference, double-antenna heading angle difference, medical scenario increases barometer observation (improves elevation accuracy); double-antenna constraint: heading angle observation equation zθ=θdual-θimu+vθ, measurement noise covariance is set to 0.01° 2 (Financial scenario) / 0.005° 2 (Medical scenario).

[0109] In the financial scenario, before the accident collision, the "high gain filtering" mode is triggered (the process noise covariance is expanded by 2 times), and the position mutation is quickly responded (such as the position update rate is increased to 20Hz when emergency braking).

[0110] In the medical scenario, the "heading lock" mechanism is enabled in the tunnel, when the double-antenna signal is stable, the heading angle update weight is forced to increase to 0.8, and the IMU drift is suppressed (the heading error is reduced from ±5° to ±1°).

[0111] In the financial scenario, when driving on the overpass, the main road / side road is accurately distinguished by the dual-antenna heading constraint (such as the 30-meter interval between the main road and the side road in Beijing's West Second Ring Road, the traditional scheme has a 20% misjudgment rate, and the present scheme reduces to 1%), ensuring the accuracy of the UBI billing mileage.

[0112] In the medical scenario, when driving on a mountainous curve, the dual-antenna real-time correction of the heading is used in combination with the IMU high-frequency sampling (200Hz) to reduce the vehicle turning angle error from ±10° to ±3°, helping the rescue system to determine whether the vehicle has deviated from the safe route (such as entering a cliff section warning).

[0113] In some embodiments, the high-frequency error and low-frequency error generated in the positioning process include: classifying the errors in the positioning process according to time characteristics, classifying the receiver clock error and signal capture noise as the high-frequency error, and estimating the high-frequency error cycle by cycle through real-time satellite signal observations; ionospheric delay, tropospheric delay are classified as low-frequency error, and low-frequency error is trend predicted according to historical observation data of reference station network and real-time weather data.

[0114] By separating errors according to time characteristics: high-frequency error (period <1 second, such as receiver clock error, multipath noise) is estimated cycle by cycle; low-frequency error (period >10 seconds, such as ionospheric slow change) is trend predicted, improving the relevance of error correction.

[0115] The error classification and estimation method includes: high-frequency error (financial scenario update rate 100Hz, medical scenario 200Hz): receiver clock error: estimated by satellite common view method, error <5ns (corresponding to 1.5m pseudorange error); signal capture noise: use adjacent epoch pseudorange difference, set 3σ threshold to remove outliers (financial scenario threshold 0.5m, medical scenario 0.3m); low-frequency error: ionospheric delay: based on 15-minute sliding average model of reference station network, generate regional ionospheric grid (grid spacing 50km, medical scenario encrypted to 20km); tropospheric delay: combined with real-time pressure / temperature data (medical scenario vehicle-mounted sensor acquisition, financial scenario call weather API), corrected by Saastamoinen model.

[0116] In the financial scenario, the ionospheric mutation across the time period (such as the delay change rate >10TECU / min at sunrise) is monitored, triggering the re-estimation of the low-frequency error to ensure the consistency of the trajectory throughout the day;

[0117] In the medical scenario, in severe weather such as heavy rain, the tropospheric correction weight is increased to 0.6 (default 0.4), and the real-time data of the vehicle-mounted barometer (accuracy ±0.5hPa) is used to reduce the elevation error from 5 meters to 2 meters.

[0118] In the financial scenario, in the cross-day accident handling, through the low-frequency error trend prediction, the trajectory drift caused by the ionospheric diurnal change is corrected (such as the error from 8 meters to 1.5 meters at 2 o'clock in the morning), and a continuous and accurate coordinate sequence across time is provided for accident liability determination.

[0119] In the medical scenario, in the cloudy weather in the mountainous area, the troposphere delay change is estimated in real time (updated every 20 seconds), the positioning height error caused by clouds and fog is avoided (the traditional scheme height error > 10 meters, and the scheme < 3 meters), and the judgment of the height of the rescue helicopter is ensured.

[0120] In some embodiments, the fusion positioning data is corrected according to the high-frequency error and the low-frequency error, to generate a vehicle positioning result according to the corrected fusion positioning data, including: the high-frequency error is filtered and corrected by using a real-time Kalman filtering algorithm, and is compensated into position, speed and heading parameters corresponding to the fusion positioning data; for the low-frequency error, the low-frequency error correction information of the regular grid point broadcast by the reference station network is received, the correction value of the adjacent grid point is matched according to the real-time position of the vehicle, the ionospheric and tropospheric delay error in the fusion positioning data is weighted and corrected, and the vehicle positioning result containing position coordinates and heading angle is generated.

[0121] The high-frequency error is compensated in real time by Kalman filtering, and the low-frequency error is weighted and corrected by using the grid correction information of the reference station network, so that the error is accurately suppressed in the dynamic scene.

[0122] The high-frequency error correction (100Hz real-time processing) includes: the state vector contains the high-frequency error term (clock drift rate, IMU zero offset), and the Kalman filtering gain matrix is dynamically adjusted: in the financial scenario, the position error covariance is increased (gain + 30%) when the acceleration is increased, and in the medical scenario, the inertial navigation weight is increased (gain + 50%) before the signal loss; the correction formula includes: , wherein K is the Kalman gain, which is compensated into the position, speed and heading parameters in real time.

[0123] The low-frequency error correction (1Hz grid matching) includes: the reference station network publishes the low-frequency error of the grid point (ionosphere / troposphere), and the grid resolution is 50km*50km in the financial scenario and 20km*20km in the medical scenario; the vehicle real-time position matches the adjacent four grid points, and the correction value is calculated by using the bilinear interpolation: , wherein wi is the distance weighting coefficient (the altitude weight is increased in the medical scenario).

[0124] In the financial scenario, when reconstructing the insurance claim trajectory, the long-term drift (such as 1-hour cumulative error) across city areas is reduced from 20 meters to within 3 meters through low-frequency error grid correction, ensuring the global consistency of positioning results on different road segments and avoiding disputes in liability determination caused by error accumulation.

[0125] In the medical scenario, in emergency rescue dispatch, high-frequency correction suppresses IMU noise caused by vehicle jolting (such as reducing speed error from 2 m / s to 0.5 m / s when emergency braking), and low-frequency grid correction improves real-time positioning accuracy to planar error <1.5 meters and height error <2.5 meters, meeting the coordinate requirements of helicopter precision air-drop rescue supplies.

[0126] In some embodiments, by addressing the feature missing problem when the Beidou signal is blocked, a signal generation adversarial network (GAN) is constructed, the generator restores weak signal phase / amplitude features, the discriminator distinguishes between real signals and generated signals, and the signal capture success rate in low signal-to-noise ratio (<-165 dBm) scenarios is improved.

[0127] The network architecture design includes: generator (G): 3 layers of transpose convolution (input layer 100-dimensional random noise → output layer generated I / Q baseband signal, size 1024x2), activation function uses ReLU (last layer tanh normalized to [-1,1]); discriminator (D): 3 layers of convolution (input real / generated signal → output 0-1 probability value), loss function is WGAN-GP (Wasserstein distance + gradient penalty), training data comes from real- collected -160~ -170 dBm weak signal samples (10,000 groups collected in each financial / medical scenario).

[0128] The signal enhancement process includes: preprocessing: downconvert the original signal to baseband and extract a 1ms data segment as input (take B1C frequency in the financial scenario and B2b frequency in the medical scenario); generate enhancement: when the measured signal-to-noise ratio is <-160 dBm, trigger GAN to generate a compensation signal, and superimpose it with the original signal according to the energy ratio of 0.3:0.7 (the financial scenario focuses on the real signal, and the medical scenario increases the weight of the generated signal to 0.5); capture verification: verify the enhanced signal through parallel code phase search, and update GAN parameters after successful capture (online incremental learning).

[0129] In the financial scenario, at the entrance of the underground parking lot (signal strength -163 dBm), the signal capture time is shortened from 800 ms in the traditional scheme to 300 ms through GAN enhancement, avoiding positioning delay caused by vehicle entry and exit, which leads to billing errors (such as the probability of missing 1 entry and exit record is reduced from 5% to 0.5%).

[0130] In the medical scene, in the dense forest area (signal strength -168 dBm), the generator preferentially restores the multipath fading characteristics of the B2b frequency point, cooperates with the inertial navigation prediction auxiliary, and improves the signal recapture success rate from 20% to 70%, ensuring the real-time position reporting of the emergency vehicle in the area without obvious road signs (delay <1 second).

[0131] In some embodiments, a deep reinforcement learning (DRL) model is constructed, with signal carrier-to-noise ratio (CN0) and Doppler frequency shift rate as state input, and outputting optimal accumulation time (T) and frequency band weight (W), replacing the traditional fixed threshold strategy, to realize self-adaptive accumulation in complex dynamic scenarios.

[0132] The DRL architecture design includes: state space (S): [current CN0, CN0 change rate in the past 3 seconds, satellite elevation angle, vehicle acceleration] (4 dimensions) action space (A): accumulation time T ∈ {50, 100, 200, 300 ms}, frequency band weight W ∈ {[0.5, 0.3, 0.2], [0.3, 0.4, 0.3], [0.2, 0.2, 0.6]} (corresponding to B1C / B2a / B2b); reward function (R): successful capture R=+100, no capture and CN0 decrease R=-50, trained using PPO algorithm, experience replay pool capacity 100,000, financial / medical scene data collected in urban canyon and mountain tunnel environment respectively.

[0133] The online decision-making process includes: collecting state S every 200 ms, inputting DRL model to generate action A; after executing the action, calculating the new CN0 and the capture state as the feedback of the next state, updating the policy network (medical scene update frequency increased to 100 ms); financial scene sets "emergency brake protection" action: when the acceleration is greater than 0.5g, the T=100ms and W=[0.5, 0.3, 0.2] are forced to be selected to ensure signal stability in the accident moment; the medical scene sets "tunnel mode" priority: when GPS signal is lost, the B2b frequency band weight is preferentially selected to be the maximum (W=[0, 0, 1]).

[0134] In the financial scene, in the urban expressway frequent lane changing scene (CN0 fluctuation ±3 dB), the DRL model shortens the accumulation parameter adjustment delay from 500 ms of the traditional scheme to 150 ms, maintains the positioning update rate at 10 Hz, and ensures the coordinate point accuracy (error <1.2 meters) of the UBI insurance in the emergency acceleration / braking event.

[0135] In the medical scene, when driving in a long tunnel (>3 kilometers), the DRL model automatically adjusts the accumulation time to 300 ms (the maximum of the traditional scheme is 200 ms), and dynamically allocates the B2b frequency band weight to 0.8, which increases the number of satellite tracking from 2 in the traditional scheme to 4, avoiding the misjudgment of the vehicle position by the rescue dispatching system (such as false reporting that the vehicle has exited the tunnel).

[0136] In some embodiments, by utilizing transfer learning techniques, a global high-precision PPP model (pre-trained on IGS global station data) is transferred to regional scenarios for non-reference station scenarios, and through a small amount of local data (100 groups per scene) fine-tuning, the single-frequency positioning accuracy under complex terrain is improved.

[0137] The transfer learning framework includes: a pre-trained model: a 12-layer Transformer network is constructed based on TensorFlow, the input is satellite ephemeris parameters, ionospheric model parameters, and receiver clock bias priori, and the output is position coordinates (3D); domain adaptation: 100 groups of labeled data (RTK true value) are collected in the hilly area of the Yangtze River Delta for the financial scenario and in the Sichuan-Yunnan mountainous area for the medical scenario, the first 8 layers of the pre-trained model are frozen, and the last 4 layers are fine-tuned (learning rate 1e-4); error compensation module: a terrain height correction branch (input SRTM 90m resolution elevation data) is added, the financial scenario focuses on plane error (weight 0.7), and the medical scenario considers elevation (weight 0.5).

[0138] The inference optimization strategy includes: in the financial scenario, in the bridge / overpass scenario, "structural feature transfer" is enabled (input bridge coordinate priori library), the plane positioning error is reduced from 4 meters to 2 meters, and the position deviation caused by traditional PPP due to multipath effect is solved (such as misjudging the lane); in the medical scenario, in the area with an altitude of >2000 meters, the barometric altimeter data (accuracy ±1m) is introduced as an auxiliary input of the transfer model, and the elevation positioning error is reduced from 8 meters to 3 meters, meeting the elevation accuracy requirements of the helicopter landing point in the mountainous area.

[0139] In the financial scenario, in the Pearl River Delta dense urban cluster (high building blockage rate >40%), the non-reference station positioning accuracy of the PPP model after transfer learning is improved from 8 meters of the traditional scheme to 3 meters, ensuring the trajectory restoration accuracy of the stolen and robbed vehicles in the offline state and providing effective tracking coordinates.

[0140] In the medical scenario, in the Hengduan Mountain area (difference >1000 meters / 10 kilometers), through terrain feature transfer and barometric data fusion, the elevation error of single-frequency PPP is reduced from 15 meters to 5 meters, avoiding the height control failure of rescue helicopters due to elevation errors (such as low flight risk).

[0141] In some embodiments, by constructing a graph neural network model, satellite positioning, inertial navigation, odometry, and dual-antenna data are represented as graph nodes, and edge weights reflect data correlation (such as time synchronization and sensor accuracy), and through graph convolution operation, the optimal fusion strategy in dynamic environment is realized.

[0142] The graph structure definition includes: node features (N): satellite nodes (CN0, elevation angle, azimuth angle), IMU nodes (acceleration, angular velocity, zero offset error), odometer nodes (wheel speed, cumulative mileage), dual-antenna nodes (baseline length, heading angle variance); edge features (E): timestamp synchronization difference (<5ms for strong connection), sensor spatial position correlation (distance <2 meters for strong connection), adjacency matrix A calculates the similarity between nodes through dynamic time warping (DTW); graph convolution layer (GCN): GAT (graph attention mechanism) is used, and the financial scenario focuses on the weight of the satellite node (number of heads 4), and the medical scenario focuses on the weight of the IMU node (number of heads 6).

[0143] The fusion decision mechanism includes: state update: graph convolution is performed every 10ms, and the fused position / velocity / heading estimation value is output; abnormality detection: when the satellite node CN0 is less than -155dBm and the IMU node angular velocity variance is greater than 0.1° 2 , the "inertia dominant" mode is triggered (the IMU weight is increased to 0.6); the financial scene adds "accident feature nodes": when three-axis acceleration is greater than 0.3g, the time stamps of all sensors are forced to be synchronized (delay <1ms), and the data fusion accuracy at the moment of collision is ensured; the medical scene adds "rescue priority nodes": when an emergency signal is received, the dual-antenna node weight is increased to 0.7, and the interference of other sensor abnormal data is suppressed. In the financial scene, in the multi-sensor time unsynchronized scene (such as odometer signal delay of 10ms), the GNN model dynamically adjusts the weight through the edge feature, reduces the fusion positioning error from 2.5 meters of traditional EKF to 1.2 meters, and ensures the accuracy of key coordinate points (such as brake start point) in accident liability determination. In the medical scene, in the ambulance passing through multiple tunnels (signal intermittent scene), the GNN model automatically identifies the satellite node failure state, increases the sum of the weights of the IMU and the odometer from 0.5 to 0.8, maintains the positioning continuity (position drift <5m / min during signal loss), and provides real-time position reference for remote medical guidance.

[0144] In some embodiments, by modeling the time correlation of high-frequency errors (receiver clock difference, multipath noise) through a long short-term memory network (LSTM), the next moment error is predicted through a historical 10-second error sequence, replacing the traditional cycle-by-cycle estimation, and the correction timeliness in dynamic scenes is improved.

[0145] The LSTM model design includes: input layer: clock error residuals of the past 10 epochs (100 ms), pseudorange multipath error, carrier phase noise (3-dimensional sequence); hidden layer: 2 layers of LSTM (128 units each) + 1 layer of fully connected layer, activation function tanh, loss function MAE, 500,000 error samples are collected in urban congestion and mountain sharp curve environments for financial / medical scenarios; prediction step: financial scenario predicts future 50 ms error (update rate 20 Hz), medical scenario predicts future 20 ms error (update rate 50 Hz).

[0146] The error correction process includes: online prediction: when the vehicle acceleration is detected to be greater than 0.2g (dynamic scenario), the LSTM prediction mode is enabled, and the traditional cycle-by-cycle estimation is used as a backup; correction fusion is weighted by the predicted error and the real-time estimated error according to the weight (dynamic scenario 0.7:0.3, static scenario 0.3:0.7), the financial scenario sets an “emergency braking trigger threshold” (when the acceleration is greater than 0.5g, the prediction weight is increased to 0.9); the medical scenario adds “physiological signal correlation”: when the vehicle-mounted monitor detects abnormal patient vital signs, the LSTM model is forced to use the highest prediction frequency (100Hz) to ensure that the positioning update is synchronized with the rescue response.

[0147] In the financial scenario, in the urban congestion frequent start-stop scenario, the LSTM model reduces the receiver clock error prediction error from 8ns in the traditional scheme to 3ns (corresponding to a pseudorange error of 2.4 meters→0.9 meters), solving the position jump problem caused by clock error mutation when following a vehicle (such as reducing the number of misjudged lane changes by 30%).

[0148] In the medical scenario, in the mountain sharp curve section (curvature radius < 50 meters), through LSTM prediction of multipath noise, the heading angle fluctuation amplitude is reduced from ±3° to ±1.5°, and combined with inertial navigation correction, the real-time heading accuracy of the ambulance on the narrow mountain road is ensured (error < 2°), avoiding navigation system misjudgment of driving direction (such as misjudging as reverse driving).

[0149] In some embodiments, by constructing a federated learning framework, multiple reference stations and vehicle-mounted terminals (financial / medical terminals are independently networked) in the region are combined to dynamically update the low-frequency error (ionosphere / troposphere) grid model under the premise of protecting data privacy, solving the delay problem of traditional centralized model updates.

[0150] The federal learning architecture includes: server side: maintaining a global low-frequency error grid model (initialized as an IGS regional model), a financial / medical scene independent server, and preventing data cross; client side: vehicle terminal uploads local observation residual (desensitization processing, only contains latitude and longitude + residual mean) regularly (financial scene every 10 minutes, medical scene every 5 minutes), without transmitting original signal data; model aggregation: using FedAvg algorithm, financial scene weight focuses on urban area client (60% of the proportion), medical scene weight focuses on suburban / mountain area client (70% of the proportion), and the aggregation period is 10 rounds.

[0151] In the financial scene, in the business-intensive area (such as Shanghai Lujiazui), through high-frequency client data (1000+ vehicles / region), the ionospheric grid resolution is improved from 50km to 10km, and the corrected plane error is reduced from 4m to 1.8m, ensuring accurate differentiation of vehicle positions in parking lots (such as adjacent parking spaces with a spacing of 2.5m, and a positioning error of <1m);

[0152] In the medical scene, a micro-federal network (50+ ambulances + 3 temporary reference stations) is established in the mountain area, when heavy rain is detected, emergency aggregation is triggered (the period is shortened to 1 minute), and a tropospheric correction grid of the heavy rain affected area is dynamically generated, reducing the height error from 10m to 4m, and ensuring the accuracy of air-drop of rescue materials in the mountain area.

[0153] In the financial scene, the federal learning model updates the ionospheric grid of high-density urban areas every hour, solves the problem of traditional RTCM broadcast delay (>30s), makes the real-time mileage statistical error of UBI insurance <0.1%, and improves the public credibility of the billing system.

[0154] In the medical scene, in the isolation area, the vehicle terminal uploads anonymized positioning residual through federal learning, assists the disease control center in dynamically updating the low-frequency error model in the region, ensures the positioning accuracy of ambulances in the isolation area is not affected by changes in the electromagnetic environment (such as temporary interference of 5G base stations), and ensures the real-time planning accuracy of the transfer route.

[0155] Please refer to Figure 4 as shown, Figure 4 is a structure diagram of a vehicle-mounted positioning device 200 based on Beidou satellite signals provided by the embodiments of the present application. The vehicle-mounted positioning device 200 based on Beidou satellite signals is used to execute the steps of the vehicle-mounted positioning method based on Beidou satellite signals shown in each of the above embodiments. The vehicle-mounted positioning device 200 based on Beidou satellite signals can be a single server or a server cluster, or the vehicle-mounted positioning device 200 based on Beidou satellite signals can be a terminal, which can be a handheld terminal, a notebook computer, a wearable device, or a robot, etc.

[0156] As Figure 4As shown, the vehicle-mounted positioning device 200 based on Beidou satellite signals comprises:

[0157] The signal acquisition unit 201 is configured to acquire Beidou satellite signals and perform multi-frequency non-coherent accumulation processing on the acquired Beidou satellite signals to enhance weak signal acquisition.

[0158] The data acquisition unit 202 is configured to acquire inertial navigation data collected by an inertial navigation device carried by the vehicle, mileage data collected by a wheel odometer, and directional data collected by a dual-antenna module.

[0159] The fusion output unit 203 is configured to enter a reference station positioning mode if a reference station signal is received, process the Beidou satellite signals according to a single-frequency real-time differential technology, acquire satellite positioning data, input the satellite positioning data, the inertial navigation data, the mileage data, and the directional data into a preset tightly coupled positioning model for fusion processing, and output fusion positioning data.

[0160] The result generation unit 204 is configured to acquire high-frequency errors and low-frequency errors generated in the positioning process, correct the fusion positioning data according to the high-frequency errors and the low-frequency errors, and generate a vehicle positioning result according to the corrected fusion positioning data.

[0161] It should be noted that, for the convenience and brevity of description, the specific working processes of the vehicle-mounted positioning device based on Beidou satellite signals and the modules described above can refer to the corresponding processes in the vehicle-mounted positioning method embodiments based on Beidou satellite signals described above, which will not be described here.

[0162] The vehicle-mounted positioning method based on Beidou satellite signals described above can be implemented in the form of a computer program, which can run on the device as shown. Figure 4

[0163] Please refer to Figure 5 , Figure 5 is a structural schematic block diagram of a computer device provided by the embodiments of the present application. The computer device comprises a processor, a memory, and a network interface connected through a device bus, wherein the memory can comprise a storage medium and an internal memory.

[0164] The storage medium can store an operating device and a computer program. The computer program comprises program instructions, which, when executed, can cause the processor to perform any kind of vehicle-mounted positioning method based on Beidou satellite signals.

[0165] The processor is configured to provide computing and control capabilities to support the operation of the entire computer device.

[0166] ​The internal memory provides an environment for the running of a computer program in a non-volatile storage medium, which, when executed by the processor, enables the processor to perform any kind of vehicle-mounted positioning method based on Beidou satellite signals.

[0167] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that, Figure 5 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the terminal to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0168] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0169] In one embodiment, the processor is configured to run a computer program stored in the memory to perform the following steps:

[0170] The Beidou satellite signals are acquired, and the acquired Beidou satellite signals are subjected to multi-frequency non-coherent accumulation processing to enhance weak signal acquisition.

[0171] Inertial navigation data collected by an inertial navigation device carried by the vehicle, mileage data collected by a wheel odometer, and directional data collected by a dual-antenna module are acquired.

[0172] If a reference station signal is received, a reference station positioning mode is entered, the Beidou satellite signals are processed according to a single-frequency real-time kinematic differential technology to acquire satellite positioning data, the satellite positioning data, the inertial navigation data, the mileage data, and the directional data are input into a preset tightly coupled positioning model for fusion processing, and fusion positioning data is output.

[0173] High-frequency errors and low-frequency errors generated during the positioning process are acquired, the fusion positioning data is corrected according to the high-frequency errors and the low-frequency errors, and a vehicle positioning result is generated according to the corrected fusion positioning data.

[0174] In some embodiments, the Beidou satellite signal includes a Beidou-3 new frequency signal; the multi-frequency non-coherent accumulation processing of the acquired Beidou satellite signal to enhance weak signal acquisition includes: processing the Beidou-3 new frequency signal according to the corresponding Beidou-3 new frequency point, performing energy accumulation on the Beidou-3 new frequency signal in each frequency segment, and dynamically adjusting the accumulation time according to the signal strength to obtain an accumulation result; and performing weighted merging on the accumulation result of each frequency segment to obtain an enhanced Beidou satellite signal.

[0175] In some embodiments, before the inertial navigation data collected by the inertial navigation device carried by the vehicle, the mileage data collected by the wheel odometer, and the directional data collected by the dual-antenna module are acquired, the method further comprises: suppressing electromagnetic interference on the Beidou satellite signal through a preset adaptive narrowband interference suppression circuit, identifying the frequency range of narrowband interference by monitoring the signal spectrum in real time, and dynamically generating a notch filter corresponding to the interference frequency; and filtering the Beidou satellite signal through the notch filter to remove the narrowband interference component.

[0176] In some embodiments, before the satellite positioning data, the inertial navigation data, the mileage data, and the directional data are input into a preset tightly coupled positioning model for fusion processing, the method further comprises: if no reference station signal is received, entering a reference station-free positioning mode to process the Beidou satellite signal using a single-frequency precise point positioning technology to obtain satellite positioning data; and inputting the satellite positioning data, the inertial navigation data, the mileage data, and the directional data into the tightly coupled positioning model.

[0177] In some embodiments, the inputting of the satellite positioning data, the inertial navigation data, the mileage data, and the directional data into the preset tightly coupled positioning model for fusion processing to output fusion positioning data comprises: establishing a state space model of the satellite positioning data, the inertial navigation data, the mileage data, and the directional data in the tightly coupled positioning model; obtaining the speed and heading corresponding to the inertial navigation data; predicting the vehicle position, the speed, and the heading as state variables and the position information corresponding to the satellite positioning data and the displacement information corresponding to the mileage data as observation variables, iteratively updating the state variables through a Kalman filtering algorithm, and introducing the dual-antenna directional data to correct and constrain the heading angle, and outputting the fusion positioning data.

[0178] In some embodiments, the high-frequency error and the low-frequency error generated in the positioning process are obtained by classifying the errors in the positioning process according to time characteristics, classifying receiver clock error and signal capture noise as the high-frequency error, and performing cycle-by-cycle estimation on the high-frequency error through real-time satellite signal observation values; and classifying ionospheric delay and tropospheric delay as the low-frequency error, and performing trend prediction on the low-frequency error according to historical observation data of the reference station network and real-time meteorological data.

[0179] In some embodiments, the fusion positioning data is corrected according to the high-frequency error and the low-frequency error, so as to generate a vehicle positioning result according to the corrected fusion positioning data, including: filtering and correcting the high-frequency error by using a real-time Kalman filtering algorithm, and compensating into position, speed and heading parameters corresponding to the fusion positioning data; for the low-frequency error, the correction information of the regular grid point low-frequency error broadcast by the reference station network is received, the correction value of the adjacent grid point is matched according to the real-time position of the vehicle, the ionospheric and tropospheric delay error in the fusion positioning data is weighted and corrected, and the vehicle positioning result containing position coordinates and heading angle is generated.

[0180] The application also provides a computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to make the processor implement the steps of the vehicle positioning method based on the Beidou satellite signal according to the first aspect.

[0181] The computer readable storage medium can be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device. The computer readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.

[0182] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A vehicle positioning method based on Beidou satellite signals, characterized in that, The method comprises the following steps: Obtaining Beidou satellite signals, and performing multi-frequency non-coherent accumulation processing on the obtained Beidou satellite signals to enhance weak signal acquisition; The non-coherent accumulation algorithm accumulates the energy of signals of different frequency bands of the same satellite, and the expression corresponding to the Beidou satellite signals comprises: ； wherein A i and B i are the amplitude values of the orthogonal carriers of each frequency band, S acc is the accumulated Beidou satellite signal, and n is the number of frequency bands. The number of accumulations is dynamically adjusted according to the scene, including fixed 8 accumulations in a financial scene and 4 fast accumulations in a medical scene for real-time requirements. A sliding window mechanism is introduced to update the accumulation window every 50 ms to avoid phase ambiguity caused by signal Doppler shift. Obtaining inertial navigation data collected by an inertial navigation device carried by the vehicle, mileage data collected by a wheel odometer, and directional data collected by a dual-antenna module; If a reference station signal is received, entering a reference station positioning mode, processing the Beidou satellite signals according to a single-frequency real-time differential technology to obtain satellite positioning data, inputting the satellite positioning data, the inertial navigation data, the mileage data, and the directional data into a preset tight combination positioning model for fusion processing to output fusion positioning data, comprising: establishing a state space model of the satellite positioning data, the inertial navigation data, the mileage data, and the directional data in the tight combination positioning model; obtaining a speed and a heading corresponding to the inertial navigation data; predicting a vehicle position according to the inertial navigation data, wherein the speed and the heading are taken as state variables, and position information corresponding to the satellite positioning data and displacement information corresponding to the mileage data are taken as observation variables, the state variables are iteratively updated through a Kalman filtering algorithm, and the heading angle is corrected and constrained by introducing the dual-antenna directional data, and the fusion positioning data is outputted; Obtaining high-frequency errors and low-frequency errors generated in the positioning process to correct the fusion positioning data according to the high-frequency errors and the low-frequency errors, and generating a vehicle positioning result according to the corrected fusion positioning data.

2. The method of claim 1, wherein, The Beidou satellite signals comprise Beidou-3 new frequency point signals; the multi-frequency non-coherent accumulation processing on the obtained Beidou satellite signals to enhance weak signal acquisition comprises: The Beidou-3 new frequency point signals are processed according to corresponding Beidou-3 new frequency points, the energy of the Beidou-3 new frequency point signals in each frequency band is accumulated, and the accumulation time is dynamically adjusted according to the signal strength to obtain an accumulation result; The accumulation results of each frequency band are weighted and combined to obtain enhanced Beidou satellite signals.

3. The method of claim 1, wherein, Before the inertial navigation data collected by the inertial navigation device carried by the vehicle, the mileage data collected by the wheel odometer, and the directional data collected by the dual-antenna module are obtained, the method further comprises the following steps: Performing electromagnetic interference suppression on the Beidou satellite signals through a preset adaptive narrowband interference suppression circuit, identifying the frequency range of narrowband interference by monitoring the signal spectrum in real time, and dynamically generating a notch filter corresponding to the interference frequency; Filtering the Beidou satellite signals through the notch filter to remove the narrowband interference components.

4. The method of claim 1, wherein, Before the satellite positioning data, the inertial navigation data, the mileage data, and the directional data are inputted into the preset tight combination positioning model for fusion processing, the method further comprises the following steps: If no reference station signal is received, entering a reference station-free positioning mode, processing the Beidou satellite signals by using a single-frequency precise point positioning technology to obtain satellite positioning data; and inputting the satellite positioning data, the inertial navigation data, the mileage data, and the directional data into the tight combination positioning model.

5. The method of claim 1, wherein, The high-frequency errors and the low-frequency errors generated in the positioning process comprise: The errors in the positioning process are classified according to time characteristics, the receiver clock difference and signal acquisition noise are classified as the high-frequency errors, and the high-frequency errors are estimated cycle by cycle through satellite signal observation values collected in real time; The ionospheric delay and tropospheric delay are classified as the low-frequency errors, and the low-frequency errors are trend predicted according to historical observation data of the reference station network and real-time meteorological data.

6. The method of claim 1, wherein, The fusion positioning data is corrected according to the high-frequency errors and the low-frequency errors, and a vehicle positioning result is generated according to the corrected fusion positioning data, including: The high-frequency errors are filtered and corrected by using a real-time Kalman filtering algorithm, and are compensated into position, speed and heading parameters corresponding to the fusion positioning data; For the low-frequency errors, low-frequency error correction information of a regular grid point broadcast by the reference station network is received, a correction value of a neighboring grid point is matched according to a real-time position of the vehicle, ionospheric and tropospheric delay errors in the fusion positioning data are weighted and corrected, and the vehicle positioning result including position coordinates and a heading angle is generated.

7. A vehicle-mounted positioning device based on Beidou satellite signals, characterized in that, including: The signal acquisition unit is configured to acquire Beidou satellite signals, and perform multi-frequency non-coherent accumulation processing on the acquired Beidou satellite signals to enhance weak signal acquisition. The non-coherent accumulation algorithm accumulates energy of signals of different frequency bands of the same satellite, and an expression corresponding to the Beidou satellite signals includes: ； wherein A i and B i are the amplitude values of the orthogonal carriers of each frequency band, S acc is the accumulated Beidou satellite signal, and n is the number of frequency bands. The number of accumulations is dynamically adjusted according to the scene, including fixed 8 accumulations in a financial scene and 4 fast accumulations in a medical scene for real-time requirements. A sliding window mechanism is introduced to update the accumulation window every 50 ms, thereby avoiding phase ambiguity caused by signal Doppler shift. The data acquisition unit is configured to acquire inertial navigation data collected by an inertial navigation device carried by the vehicle, mileage data collected by a wheel odometer, and directional data collected by a dual-antenna module. The fusion output unit is configured to enter a reference station positioning mode if the reference station signal is received, process the Beidou satellite signals according to a single-frequency real-time differential technology, acquire satellite positioning data, input the satellite positioning data, the inertial navigation data, the mileage data and the directional data into a preset tightly coupled positioning model for fusion processing, and output fusion positioning data, including: establishing a state space model of the satellite positioning data, the inertial navigation data, the mileage data and the directional data in the tightly coupled positioning model; acquiring speed and heading corresponding to the inertial navigation data; predicting a vehicle position according to the inertial navigation data, wherein the speed and the heading are used as state variables, and position information corresponding to the satellite positioning data and displacement information corresponding to the mileage data are used as observation variables, the state variables are iteratively updated through a Kalman filtering algorithm, and a heading angle is corrected and constrained by introducing the directional data of the dual-antenna, and the fusion positioning data is output. The result generation unit is configured to acquire high-frequency errors and low-frequency errors generated in a positioning process, correct the fusion positioning data according to the high-frequency errors and the low-frequency errors, and generate a vehicle positioning result according to the corrected fusion positioning data.

8. A computer device, comprising: The computer device includes a memory and a processor; The memory is configured to store a computer program; The processor is configured to execute the computer program and implement the method in any one of claims 1 to 6 when the computer program is executed.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program causes the processor to implement the method in any one of claims 1 to 6 when the computer program is executed by the processor.

Images