Multi-source navigation signal switching and fusing method and navigation system thereof

Abstract

The invention discloses a multi-source navigation signal switching and fusion method and a navigation system thereof, and the method comprises the steps: carrying out the quality evaluation of signals from different navigation source GNSS, and determining a main navigation source based on a quality evaluation result; when the quality score of the main navigation source is lower than a preset threshold value or the GNSS observation information meets a preset switching condition, switching between the main navigation source and the standby navigation source is carried out, and smoothing processing is carried out on the new navigation solution and the old navigation solution in the switching process of the main navigation source and the standby navigation source; and performing multi-source data fusion and state estimation on the data of the navigation source GNSS, the inertial measurement unit, the geomagnetic sensor and the air pressure sensor through a filtering algorithm, and outputting a navigation result. According to the navigation system, a plurality of multi-constellation GNSS receiving chips, an inertial measurement unit, a geomagnetic sensor, an air pressure sensor and a micro-control unit are vertically interconnected in a single package through a wafer-level heterogeneous integration process. According to the invention, the positioning precision and robustness in a complex environment can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of navigation positioning, in particular to a multi-source navigation signal switching and fusion method and a navigation system thereof, and belongs to the cross-technology category of navigation receiver design, sensor fusion and microsystem packaging integration. BACKGROUND

[0002] Global Navigation Satellite System (GNSS) can provide high-precision positioning in open environments, but in urban canyons, tunnels, underground spaces and interference areas with high-rise buildings, the pseudorange / carrying wave observation is significantly biased due to multipath and NLOS (non-line-of-sight) propagation, and the positioning accuracy and continuity are significantly degraded, and even the lock is lost. Recent research quantifies the systematic destruction and uncertainty sources of urban environment on GNSS stability from visible star prediction, pedestrian / mobile positioning, etc. [1–2] [4]

[0003] To alleviate the inherent deficiencies of single constellation / single frequency in redundancy and geometric conditions, the industry gradually adopts multi-constellation / multi-frequency and multi-sensor fusion to improve availability, DOP, integrity and robustness. Simulation and experimental evaluation shows that multi-constellation parallel observation can significantly improve the robustness of "visible satellite number + geometric distribution", thereby improving the continuity and convergence in restricted environments such as urban canyons [3-4] At the same time, GNSS / IMU combined navigation increasingly relies on integrity monitoring (RAIM / ARAIM) and efficient identification and rejection of abnormal observations in complex environments - the algorithm update, parameter sensitivity and architecture report of ARAIM show that autonomous alarm and rapid failure detection on the receiver side are the key paths to ensure integrity in multi-constellation and complex fault scenarios [5-7]

[0004] However, in the GNSS weak signal or lock loss stage, the navigation system often faces the stability problem of main and backup source switching: if only fixed thresholds or single weighting are relied on, switching sudden jumps or filter divergence are easy to occur. For this reason, some research has introduced fuzzy logic / adaptive mechanism (such as adaptive Kalman, fuzzy adaptive + strong tracking EKF, etc.) in GPS / INS fusion, and according to the quality indicators such as SNR, number of visible stars, PDOP, etc., the weight / noise covariance is adaptively adjusted, which significantly improves the stability and switching performance in high dynamic and weak signal conditions [8–9] At the same time, the vertical and attitude dimensions are particularly vulnerable when GNSS degrades, and barometric altitude and geomagnetic heading have been proven to provide effective constraints for INS, reducing vertical drift and improving availability and height stability in long weak or invisible conditions [10–11]

[0005] ​​​​At the hardware level, traditional systems are mostly assembled at the board / module level, making it difficult to simultaneously address size, interconnect parasitics, and EMI shielding. In recent years, wafer-level heterogeneous integration (WLP / 2.5D / 3D-HI) technology has enabled high-density vertical interconnection of multiple chips and processes within a single package, shortening interconnection paths, reducing parasitic effects, and achieving EMI isolation and unified clock / time synchronization at the package level. The development of MEMS / CMOS co-process technology has made it feasible to integrate inertial measurement units, Lorentz force-type MEMS geomagnetic sensors, and barometric pressure sensors with GNSS receiver chips within the same package, thereby enabling integrated deployment of multi-source sensing and fusion computing on miniaturized platforms. [12–13] .

[0006] The shortcomings of existing technologies are as follows: Although the literature has discussed technologies such as multi-constellation GNSS / INS fusion, integrity monitoring and anomaly rejection, barometric and geomagnetic auxiliary constraints, and wafer-level heterogeneous integration, there is still a lack of a systematic engineering solution that integrates multi-constellation GNSS + IMU + geomagnetic + barometric + MCU into a single wafer-level package, and is equipped with a closed-loop algorithm of "quality assessment → intelligent master / slave switching → smooth transition → robust fusion (anomaly rejection + dynamic weighting)" to simultaneously meet the dual requirements of "miniaturization, low power consumption, strong anti-interference" and "continuous high precision in complex environments".

[0007] [1]Zheng, S.; Zeng, K.; Li, Z.; Wang, Q.; Xie, K.; Liu, M.; Xie, S. Improving the Prediction of GNSS Satellite Visibility in Urban Canyons Based on a Graph Transformer . NAVIGATION2024, 71(4), navi.676. DOI:10.33012 / navi.676.[2] Weng D., et al. Sidewalk matching: a smartphone-based GNSSpositioning method in urban canyons. Satellite Navigation , 2025.[3] Wang, L;Groves, PD; Ziebart, MK Multi-Constellation GNSS Performance Evaluation for Urban Canyons Using Large Virtual Reality City Models . Journal ofNavigation,2012, 65(3): 459–476.DOI: 10.1017 / S0373463312000082.[4] TeunissenP.JG, Montenbruck O. (Eds.). Springer Handbook of Global Navigation Satellite Systems. Springer, 2017. [5] Blanch J., Walter T., Enge P. Baseline Advanced RAIM User Algorithm: Proposed Updates. ION ITM , 2022. [6] Lee Y., et al. Sensitivity of Advanced RAIM Performance to Integrity Support Message Parameters. NAVIGATION , 2021. [7] EU–US Working Group-C. ARAIM Technical Subgroup Milestone 2 Report (2015); Milestone 3 Report (2016). [8] Sun B., et al. An Improved Innovation Adaptive Kalman Filter for Integrated INS / GPS Navigation. Sustainability , 2022. https: / / www.mdpi.com / 2071-1050 / 14 / 18 / 11230 [9] (IET)INS / GPS Sensor Fusion based on Adaptive Fuzzy EKF with Strong Tracking Strategy, 2021. https: / / ietresearch.onlinelibrary.wiley.com / doi / 10.1049 / rsn2.12144

[10] Kim J.H., et al. A Baro-Altimeter Augmented INS / GPS Navigation System for a UAV. (SatNav 2003 conference paper) https: / / users.cecs.anu.edu.au / ~Jonghyuk.Kim / pdf / SatNav_2003_Kim.pdf

[11] Hajiyev C., et al. Integration of Barometric and GPS Altimeters via Adaptive Data Fusion Algorithm. 2020 / 2021. https: / / web.itu.edu.tr / cilden / pdfs / 2021_Integrationofbarometric andGPSaltimetersviaadaptivedatafusionalgorithm.pdf.

[12] Tu C., Ou-Yang X., Zhang X. Single-Structure 3-Axis LorentzForce Magnetometer based on an AlN-on-Si MEMS Resonator. Microsystems& Nanoengineering , 2024. https: / / www.nature.com / articles / s41378-024-00696-3.

[13] Valle JJ, et al. Design, Fabrication, Characterization and ReliabilityStudy of CMOS-MEMS Lorentz-Force Magnetometers. arXiv , 2021. https: / / arxiv.org / abs / 2109.10980. Summary of the Invention To address at least one of the problems existing in the prior art, this invention provides a method for switching and fusing multi-source navigation signals and a navigation system based on wafer-level heterogeneous integration. This method enables dynamic weighted fusion of GNSS, IMU, geomagnetic, and barometric data, maintains high accuracy and reliability in environments with multipath interference and satellite signal blockage, and improves positioning accuracy and robustness in complex environments.

[0008] This invention provides a method for switching and fusing multi-source navigation signals, comprising the following steps: The quality of GNSS signals from different navigation sources is assessed, and the current primary navigation source is determined based on the assessment results. When the quality score of the primary navigation source is lower than the preset threshold, or when the GNSS observation information meets the preset switching conditions, the primary navigation source and the backup navigation source are switched, and the old navigation solution and the new navigation solution are smoothed during the switching process. The navigation results are output by performing multi-source data fusion and state estimation on the measurement data of navigation sources GNSS, inertial measurement unit, geomagnetic sensor and barometric pressure sensor through filtering algorithm.

[0009] Furthermore, the quality assessment includes calculating the quality score of each navigation source and selecting the navigation source with the highest quality score as the current primary navigation source. The quality score of the navigation source is obtained based on each GNSS observation information and by adaptively adjusting the fusion weight.

[0010] Furthermore, the preset switching conditions satisfied by the GNSS observation information include at least one of the following: The carrier signal-to-noise ratio is below the preset lower limit; The number of visible satellites is lower than the preset value; The position accuracy factor (PDOP) exceeds the preset threshold. Carrier lock state interrupted; The cycle jump rate exceeds a preset threshold. Further, the smoothing process includes: Real-time calculation of changes in position, velocity, and heading angle between the old and new navigation solutions; When any change exceeds the preset soft constraint threshold, a gradual transition is achieved by using first-order hysteresis filtering or spline interpolation. When the change is within the threshold range, switch directly to the new main navigation source; The smoothing time constant is adaptively adjusted based on the difference in quality between the primary and backup power sources and the vehicle's mobility index to control the gradual speed of the switching process.

[0011] Furthermore, the multi-source data fusion includes: Define the state vector of the navigation system and establish the state prediction equation; Based on the real-time quality scores of each measurement source, the measurement noise covariance matrix is ​​dynamically adjusted to enable adaptive weighting of GNSS position and velocity measurements, barometric altitude measurements, and geomagnetic navigation vector measurements. The navigation state after multi-source fusion is obtained by updating through filtering.

[0012] Furthermore, the measurement models for each measurement source include: The measurement model for the GNSS position of the primary navigation source is as follows: , The GNSS position noise from the primary navigation source is assumed to be zero-mean white Gaussian noise. The measurement noise covariance matrix of the main navigation source GNSS Adaptively adjusted based on the quality score of the primary navigation source GNSS; The measurement model for the barometric altimeter is as follows: , This is barometric altitude noise, assumed to be zero-mean Gaussian white noise. , This is the noise covariance matrix for barometric altitude measurement. , This is the reference noise variance for barometric altitude measurement. To provide a real-time quality score for the barometric pressure sensor. To prevent tiny constants with a denominator of zero; The measurement model of the geomagnetic sensor is , for posture Yaw angle in the middle; This is geomagnetic heading angle noise, assumed to be zero-mean Gaussian white noise. , The noise covariance matrix of geomagnetic measurements. , This is the reference noise variance for geomagnetic airborne vector measurements. To provide a real-time quality score for the geomagnetic sensor. To prevent tiny constants with a denominator of zero.

[0013] Furthermore, it also includes the following steps: performing abnormal measurement elimination and integrity monitoring to identify and eliminate abnormal measurements; when the navigation source GNSS is in a state of degradation or loss of lock but not completely failed, using barometric altitude and geomagnetic heading as external measurements to correct the inertial navigation results in order to maintain the stability of the navigation solution. When GNSS is completely unusable, navigation results are made continuous and stable by using inertial measurement units to calculate, barometric altitude and geomagnetic heading constraints, and covariance expansion in a coordinated manner.

[0014] The present invention provides a navigation system based on multi-navigation modules and wafer-level integration. It adopts wafer-level heterogeneous integration packaging to integrate multiple multi-constellation GNSS cores and navigation processing cores in the same package in a parallel arrangement or partial stacking manner. The navigation processing core includes a microcontroller unit, an inertial measurement unit, a geomagnetic sensor and a barometric pressure sensor. The microcontroller unit is configured with a control program for implementing the aforementioned method.

[0015] The microcontroller operates a navigation signal switching and fusion method, which includes navigation source quality assessment, primary / backup switching, smooth transition, multi-source fusion and integrity monitoring, to achieve dynamic weighted fusion of GNSS, IMU, geomagnetic and barometric data.

[0016] Furthermore, the package incorporates a multi-RF channel isolation structure and a unified clock / timestamp synchronization circuit, and achieves heterogeneous chip interconnection through a shared data bus and power network within the package. It also features a package-level electromagnetic shielding structure to reduce RF interference and interconnect parasitic effects.

[0017] The package incorporates a multi-RF channel isolation structure, a unified clock synchronization circuit, and a package-level electromagnetic shielding structure to reduce RF interference and ensure time consistency of multi-source data.

[0018] Furthermore, the wafer-level heterogeneous integration employs at least one or more of the following processes to achieve vertical interconnection and high-density wiring of heterogeneous chips; and / or, the GNSS chip and the navigation processing chip are interconnected laterally via RDL or vertically via TSV.

[0019] Compared with the prior art, the present invention can achieve at least the following beneficial effects: This invention balances hardware integration and software intelligence, and is applicable to various platforms such as vehicles, aircraft, and unmanned systems. It can maintain high precision and high reliability in environments with multipath interference and satellite signal blockage.

[0020] On the hardware side, a wafer-level heterogeneous integration solution is proposed, consisting of multiple multi-constellation GNSS receiver chips, an IMU, geomagnetic sensors, barometers, and an MCU. The package achieves multi-RF front-end isolation and unified clock / time synchronization, and has the advantages of low parasitics, strong EMI shielding, and controlled power consumption.

[0021] On the software side, an adaptive link is constructed that consists of "quality scoring based on SNR / visible satellite count / PDOP → intelligent primary / backup switching → smooth transition → multi-source robust fusion (including RAIM / FDE anomaly removal and dynamic weighting)". When GNSS degrades or loses lock, barometric altitude and geomagnetic heading are introduced to constrain the inertial calculation, which significantly reduces switching delay and abrupt jump risk, and suppresses the impact of abnormal observations on fusion performance, thereby achieving high-precision and robust continuous positioning capability on a miniaturized platform. Attached Figure Description

[0022] Figure 1 This is an integrated schematic diagram of the navigation system in an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the packaging in an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram illustrating the steps of a multi-source navigation signal switching and fusion method provided in an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Global Navigation Satellite Systems (GNSS) offer high positioning accuracy in open environments, but in complex scenarios such as urban canyons, tunnels, indoor-outdoor switching, and electromagnetic interference, signal weakening, insufficient visible satellites, and multipath errors can occur, leading to decreased positioning continuity and stability. In engineering, robustness can be improved through a combination of GNSS, Inertial Measurement Unit (IMU), geomagnetic, and barometric pressure systems. However, existing solutions are mostly board-level / module-level assemblies, making it difficult to reduce size and power consumption, and interconnect parasitics and electromagnetic interference are also difficult to control. Furthermore, they support limited constellations or only a single constellation, resulting in large switching delays between primary and backup sources, and a tendency for drift and sudden jumps during weak / lost-lock phases. Existing fusion algorithms often rely on fixed thresholds or fixed weights, offering limited ability to suppress anomalous observations and degraded data.

[0027] In terms of hardware structure, this invention provides a high-precision navigation system based on multi-navigation chips and wafer-level heterogeneous integration. This high-precision navigation system employs a hybrid integration method of 2.5D parallel arrangement and partial 3D stacking, achieving high-density interconnection between multiple constellation GNSS chips (defined as GNSS-A, GNSS-B, GNSS-C, etc.) and navigation processing chips on the same package substrate. It also includes power management circuitry and an RF front-end, such as... Figure 1 and Figure 2 As shown. The navigation processing core includes: The core area of ​​the microcontroller unit (MCU) contains a storage area for algorithms such as quality assessment, master / slave switching, and fusion filtering. An inertial measurement unit (IMU) includes at least a three-axis accelerometer and a three-axis gyroscope; Geomagnetic sensors and barometric pressure sensors.

[0028] Each GNSS chip has multi-constellation receiving capability (supporting at least GPS, BeiDou, and GLONASS) and is equipped with an independent RF front-end. The package incorporates a multi-RF channel isolation structure and a unified clock / timestamp synchronization circuit to achieve parallel sampling and time consistency across different GNSS receiving links. Heterogeneous chips are interconnected via a shared data bus and power network within the package, and a package-level electromagnetic shielding structure is implemented to reduce RF interference and interconnect parasitic effects.

[0029] Multi-RF channel isolation structures are suitable for both parallel (2.5D) and partial stacking (3D-HI) layouts. These structures include metal shielding walls, ground fences, and differential trace isolation strips, along with multiple bandpass / bandstop filters and low-noise amplification links to reduce inter-channel interference. The metal shielding walls, ground fences, and differential trace isolation strips correspond to electromagnetic isolation measures at the package level, wiring level, and signal level, respectively.

[0030] Wafer-level heterogeneous integration employs at least one or more of the following processes: through-silicon vias (TSVs), microbump bonding, and redistribution layers (RDLs). High-speed data and clock distribution can be achieved between GNSS chips and navigation processing chips through lateral interconnects via redistribution layers (RDLs) or vertical interconnects via TSVs. Packaging preferably utilizes metal covers, shielding cavities, or ground ring structures to form package-level electromagnetic shielding.

[0031] The unified clock / timestamp synchronization circuit uses GNSS reference time to synchronize the IMU, geomagnetism, barometric pressure and each GNSS receiving link, so that the cross-sensor time alignment error is no greater than a preset value, such as 1ms.

[0032] Data and clock distribution are horizontally interconnected via redistribution layers (RDLs), while vertical interconnection is achieved using through-silicon vias (TSVs) in local stacked cells to reduce interconnect length and parasitic effects. A unified clock / timestamp synchronization circuit uses GNSS reference time to perform nanosecond-level time synchronization with IMUs, geomagnetic sensors, barometers, and each GNSS receiver link, ensuring that the cross-sensor time alignment error is no greater than 1 ms.

[0033] The microcontroller unit (MCU) is equipped with a control program for navigation signal quality assessment, primary / backup source switching, smooth transition, multi-source fusion, and integrity monitoring (the steps for implementing the multi-source navigation signal switching and fusion method). The control program can automatically complete the primary / backup source switching and maintain the continuity and stability of the navigation solution when the navigation signal quality deteriorates or is interrupted.

[0034] Please see Figure 3 The present invention provides a method for switching and fusing multi-source navigation signals, comprising the following steps: S1. Quality Assessment: The quality of signals from different navigation sources (GNSS chips) is assessed, and the primary navigation source is determined based on the quality assessment results.

[0035] The different navigation sources in this step are multiple GNSS cores. This step is used to evaluate and filter navigation signals from multiple GNSS cores.

[0036] The quality assessment is based on a comprehensive score of the navigation source's signal quality parameters and availability, which is used to dynamically adjust the weights of each navigation source during the fusion process. Signal quality parameters and availability refer to raw observation indicators such as carrier signal-to-noise ratio (SNR), number of visible satellites, PDOP (Potentially Target Opportunity), carrier lock-in status, and cycle slip rate. The microcontroller unit (MCU) calculates a quality score based on these indicators, which is then used to dynamically adjust the fusion weights. Initially, the MCU drives each GNSS chip to periodically acquire raw GNSS observation information, which is synchronously recorded with the health status indicators of the inertial measurement unit (IMU), geomagnetic sensor, and barometric pressure sensor. This GNSS observation information includes carrier signal-to-noise ratio (SNR), number of visible satellites, etc. Position accuracy factor Carrier-locked state and cycle slip rate .

[0037] For each navigation source, the microcontroller unit (MCU) calculates the current... Time of the first Quality score of GNSS source receiver link The quality score is calculated using the following formula: (The formula includes a function that triggers master / slave switching based on the sorting order and adaptively adjusts the fusion weights.)

[0038]

[0039] in: For the first Quality score of each GNSS source receiver link; Normalized carrier signal-to-noise ratio; Configure the maximum number of visible characters allowed; This is the normalized position precision factor; This indicates the carrier-locked state (value 1 when locked, value 0 when unlocked). Normalized cycle slip rate (the larger the value, the worse the result); For adaptive weighting, adjustments can be made based on historical statistics and exponential moving averages. , and Each is the current Time and -1 moment The GNSS source receiving link An adaptive weight, An adaptive adjustment factor for the quality score of the GNSS source-receiver link. It can be calibrated offline or fine-tuned online based on the correlation between each indicator and integrity / residual. Indicates the first The GNSS source receiving link A baseline weight for each quality metric is determined offline based on the correlation between the metric and navigation performance (such as residuals, integrity, and stability). In one embodiment...

[0040] The microcontroller unit (MCU) selects the navigation source with the highest quality score as the primary navigation source at the current moment.

[0041] = ; For the current number The source quality score at any given moment.

[0042] S2. Primary / Backup Source Switchover: Primary / backup source (primary navigation source and backup navigation source) switchover is triggered when the quality score of the primary navigation source falls below a preset threshold or when GNSS observation information meets preset conditions. Primary / backup source switchover is triggered when any of the following conditions are detected: 1. Current number Momentary Source Quality Score lower than the The next update sets the quality score threshold. Microcontroller Unit (MCU) through Formula updates the main source quality score threshold The update method can be performed at fixed time intervals.

[0043] (101) Initial threshold The operating range of GNSS signals is determined based on the stable operating range of the signals in the experimental or simulation environment. In one embodiment, =0.4~0.6, normalized interval [0,1].

[0044] (102) Each time an update is performed, first calculate the most recent At any moment (e.g.) =10) Average of the main source quality scores ] / 10, then compare that average with the first The next updated source quality score threshold is calculated using a smoothing factor. We perform weighted fusion to obtain the main source quality score threshold for this update: , In one embodiment, the average of the primary source quality scores is... Take a value of 0.1 to 0.3.

[0045] 2. The carrier signal-to-noise ratio remains consistently below the lower limit of the carrier frequency. 3. Number of visible satellites Insufficient, below a preset quantity value; in one embodiment... At that time, it was determined that the number of visible satellites was insufficient; 4. Position accuracy factor In one embodiment, the accuracy exceeds a preset threshold. The limit is exceeded at that time; 5. Carrier lock state interrupted; 6. The cycle slip rate exceeds the allowable cycle slip rate threshold. The allowable threshold depends on the frequency of carrier phase measurement and the dynamic environment. It can be adaptively updated by statistically analyzing the moving average and standard deviation of the historical average cycle slip rate. The microcontroller unit (MCU) selects the navigation source with the highest quality score as the new primary source and records the switching reason (such as "lost lock" or "high cycle jump rate") for subsequent optimization.

[0046] S3. Smooth transition: Smooth the old navigation solution and the new navigation solution during the switching process between primary and backup sources.

[0047] The smoothing process reduces abrupt changes in position, velocity, or attitude solutions (mainly heading angle differences) during the switching process, improving the continuity of navigation output. During the primary / backup source switching process, the microcontroller unit (MCU) calculates the position change (Δpos), velocity change (Δvel), and heading angle change (Δyaw) of the old and new navigation solutions in real time, representing the position, velocity, and heading angle differences before and after the switch or between two adjacent moments. If the changes exceed preset soft constraint thresholds (in one embodiment, the preset soft constraint thresholds are: Δpos ≤ 5 m, Δvel ≤ 2 m / s, Δyaw ≤ 10°), the abrupt solution is not used directly. Instead, a gradual transition is achieved through first-order hysteresis filtering or spline interpolation. When the changes are within the threshold range, the system directly switches to the new primary source to accelerate convergence.

[0048] Smoothing time constant (In one embodiment, Within the range of 0.1–5 s, the new primary source weight is calculated as follows: Perform calculations. The sampling time interval for the GNSS navigation source. The number of seconds since the start of the switchover from primary to backup source At that moment, New Source Weight , As the weight increases, the weight gradually increases exponentially. When the weight of the new primary source is close to 1, it indicates that the new primary source has completely taken over. (Smoothing time constant) Reflecting the gradual change speed during the switching process, the main control unit (MCU) adjusts the speed based on the quality difference between the primary and backup power sources. , With vehicle mobility index (Acceleration and angular velocity characteristics output by the IMU) Adaptively adjust the smoothing time constant , fusion weight The weights are smoothly increased from 0 to full weight to ensure that there are no abrupt changes or filter divergences in position, velocity, and heading angles during the switching process. The quality score of the main source before the switch. This is the quality score for the new primary source after the switch. After the switch is complete, Within the convergence window, the preset soft constraint thresholds (Δpos, Δvel, Δyaw) are continuously monitored, and the update frequency of external observations is appropriately reduced before anomaly recovery to avoid introducing abrupt solutions into the fusion filter.

[0049] Smoothing time constant Adaptive adjustment is performed using the following formula: ; In the formula, and These are the minimum and maximum smoothing time constants set for the navigation system, respectively. Used for rapid switching when the signal is stable and there is low mobility. Used to suppress sudden jumps during signal abrupt changes or high maneuvers, in one embodiment... Take 3–10 times the sampling period of the navigation system (approximately 0.3–0.5 s). Take 0.5–1 times the attitude change period or the filter settling time (approximately 3–5 seconds).

[0050] In one embodiment, the microcontroller unit (MCU) can perform first-order hysteresis filtering or spline interpolation smoothing on the navigation solution.

[0051] S4. Multi-source fusion: Fusion of data from the main navigation source GNSS, inertial measurement unit, geomagnetic sensor and barometric pressure sensor to output navigation solution.

[0052] Multi-source fusion achieves state estimation based on filtering algorithms, and can be based on the quality score of the main navigation source. Dynamically adjust the global measurement noise covariance matrix When the primary navigation source signal quality is high, reduce the measurement noise parameter of the corresponding navigation source. This enhances the observational impact of the navigation source during the filtering and update phase; when the navigation signal quality is poor, the measurement noise parameter is increased to reduce its role in the fusion calculation. The parameter adjustment can be performed linearly, i.e., quantitative... The measured noise parameters change linearly inversely with the quality score, thus achieving a smooth and continuous adaptive weighting process.

[0053] In one embodiment, the filtering algorithm employs an extended Kalman filter; it is understood that the filtering algorithm is not limited to this one. For Extended Kalman Filter (EKF), other algorithms can also be used. For example, when the navigation system has significant nonlinearity or non-Gaussianity, equivalent algorithms such as Unscented Kalman Filter (UKF) and Particle Filter can be used instead.

[0054] In one embodiment, the filtering algorithm employs an extended Kalman filter (EKF). Within the EKF framework, the inertial estimation results from the inertial measurement unit are used as the state prediction input. This is combined with dynamic weighted fusion of GNSS position and velocity measurements, barometric altitude measurements, and geomagnetic aerobatic measurements from the main navigation source to output a high-precision position vector. velocity vector with posture The measurement noise covariance of the filter is dynamically adjusted based on the real-time quality scores of each measurement source to create a global measurement noise covariance matrix. This achieves weighted enhancement of high-quality observations and suppression of low-quality observations, thereby obtaining a stable and continuous fusion navigation solution. The specific fusion steps are as follows: 1) Perform state prediction (INS estimation) The state vector of a navigation system is defined as:

[0055] in: , is the position vector, that is, the carrier in The position of the time in the North-East-Down coordinate system (NED). These are the displacements of the carrier in the North, East, and Down directions, respectively. That is, the carrier in the northeast coordinate system Speed ​​at any moment These are the velocity components along the north, east, and ground directions, respectively; That is, the carrier is Euler angles of the moment These are roll, pitch, and yaw. , and The rotation angle of the shaft;

[0056]

[0057] 2) Establish the state prediction equation (discrete form):

[0058] in, To predict the state vector, For process noise, Let covariance matrix be the variance matrix. =[ , ], The nonlinear state prediction model is given by the inertial navigation equation (INS). Based on the well-known inertial navigation calculation principle, the model can calculate the carrier's attitude, velocity and position changes from the triaxial acceleration and angular velocity measured by the inertial measurement unit. The current moment; The acceleration vector measured by the IMU is one of the inputs for state prediction. The three-axis angular velocity vector measured by the IMU is the second input for state prediction. 3) Establish a measurement model (integrating GNSS, air pressure, and geomagnetism), and use external observation information to correct inertial navigation errors in real time to achieve multi-source information fusion.

[0059] The Inertial Measurement Unit (IMU) is a state prediction input and not an external measurement source, therefore it is not included in the measurement model. The measurement model only includes GNSS / barometry / geomagnetic data and is used to correct the state predictions derived from the IMU. After establishing the measurement models for GNSS, barometry, and geomagnetism, the filter dynamically adjusts the corresponding measurement noise covariance matrix based on the real-time quality scores of each measurement source. This enables adaptive weighting of GNSS position and velocity measurements, barometry and altitude measurements, and geomagnetic navigation vector measurements during the fusion and update phase, updating the posterior state estimate after multi-source fusion. This method achieves continuous and robust fusion navigation results by introducing quality-score-based weight adjustment into the measurement noise covariance matrix, thereby enhancing high-quality observations and suppressing low-quality observations.

[0060] (301) Measurement model of the GNSS position of the main navigation source , in, The GNSS position noise from the primary navigation source is assumed to be zero-mean white Gaussian noise. The measurement noise covariance matrix of the main navigation source GNSS Adaptively adjusted based on the quality score of the primary navigation source GNSS: = , The quality score of the primary navigation source GNSS is between (0, 1). This represents the nominal measurement noise level corresponding to a quality score of 1 (i.e., full score) for the primary navigation source GNSS. It reflects the variance or covariance of the reference position / velocity measurement under normal, unobstructed, and excellent signal quality conditions. It is a 3×3 diagonal array, and the reference noise for horizontal position measurement can be taken as (N / E): The reference noise for vertical position measurement can be taken as (N / E): In one embodiment, .

[0061] (302) Measurement model of barometric pressure sensor

[0062] in, For position vector The third component The height obtained from the conversion; This is barometric altitude noise, assumed to be zero-mean Gaussian white noise. , This is the noise covariance matrix for barometric altitude measurement, used to characterize the uncertainty of barometric altitude measurement under the current pressure environment, and can be adaptively adjusted according to pressure stability. Specifically, the microcontroller unit (MCU) calculates the real-time quality score of the barometric pressure sensor during the quality assessment phase. This quality score comprehensively reflects the stability, continuity, and short-term pressure interference of the barometric pressure signal. For example, it can be obtained through normalization and weighted summation based on indicators such as the variance of the barometric altitude time series, the magnitude of the residual with the inertial navigation system's predicted altitude, and whether the barometric pressure signal exhibits jumps. During measurement updates, the filter adaptively adjusts the barometric altitude measurement noise covariance based on the real-time quality score of the barometric pressure sensor. ,in, The reference noise variance for barometric altitude measurement can be taken as (0.5m) in one embodiment. 2 , To prevent small constants with a denominator of zero (e.g., 10) -3 When the atmospheric pressure environment is stable and the disturbance is weak, High, near The weight of air pressure measurement is increased accordingly when air pressure fluctuations are significant, or when there are short-term pressure disturbances or oscillations. Decrease Correspondingly, the filter automatically reduces the weight of barometric pressure measurements to suppress the impact of abnormal barometric pressure measurements on altitude estimation, for example: In one embodiment,

[0063] , , These are the minimum, median, and maximum values ​​of the noise covariance in barometric pressure measurement, respectively.

[0064] (303) Measurement model of geomagnetic sensor

[0065] in, for posture Yaw angle in the middle; This is geomagnetic heading angle noise, assumed to be zero-mean Gaussian white noise. , The geomagnetic measurement noise covariance matrix characterizes the uncertainty of geomagnetic observations under the current magnetic environment. Its magnitude can be adaptively adjusted according to the level of geomagnetic interference. Specifically, the microcontroller unit (MCU) calculates the real-time quality score of the geomagnetic sensor during the quality assessment phase. The real-time quality score of this geomagnetic sensor comprehensively reflects the stability, continuity, and degree of magnetic field interference of the geomagnetic signal. For example, it can be obtained by weighting normalized indices such as the deviation of the geomagnetic field modulus from its nominal value, the variance of the flight vector measurements over a period of time, and the residuals between the geomagnetic heading and inertial heading. During measurement updates, the filter dynamically adjusts the geomagnetic measurement noise covariance based on the real-time quality score of the geomagnetic sensor. ,in, The reference noise variance for geomagnetic airborne vector measurements, in one embodiment, can be taken as... ; To prevent the use of tiny constants with a denominator of zero, in one embodiment, the value is taken as... When the local magnetic environment is stable and interference is weak, High, near The weight of geomagnetic measurements increases; when geomagnetic interference intensifies or magnetic field fluctuations become significant. Decrease Correspondingly, the filter automatically reduces the weight of geomagnetic measurements to suppress the impact of anomalous magnetic interference on attitude estimation. In one embodiment, the geomagnetic headway measurement noise covariance... Real-time quality scoring based on geomagnetic sensors Segmented value retrieval:

[0066] , , These are the minimum, median, and maximum values ​​of the geomagnetic airborne vector measurement noise covariance.

[0067] 4) Kalman filter update Calculate the predicted residuals: , For the first The time is a unified observation vector formed by the combination of GNSS, air pressure, and geomagnetic observations. Indicates the first The state prediction vector (position, velocity, attitude) of the real-time navigation system. For transpose, For the observation model function of the navigation system; Update the global measurement noise covariance matrix =[ Furthermore, it adaptively adjusts based on the measurement residuals:

[0068] in, For the global measurement residual vector, This is a smoothing coefficient, which can be selected according to the environmental characteristics of the navigation system (e.g., 0.7-0.9). (Superscript) Indicates transpose; Kalman gain:

[0069] Status Update:

[0070] The covariance update formula is:

[0071] in, and These are the predicted and updated state covariance matrices, respectively. For unit array; For the measurement matrix, for Posterior state estimation at time 1.

[0072] Updated state covariance matrix This serves as the initial covariance for the state prediction in the next sampling period and is used to calculate the Kalman gain and the uncertainty of the navigation solution at the next time step. During the state prediction phase, it utilizes... The error covariance is propagated, where For state prediction models For the Jacobian matrix of the state vector, in Evaluate at the specified location; The process noise covariance matrix for the prediction phase is determined by the angular velocity noise density, acceleration noise density, and zero-bias random walk characteristics of the inertial measurement unit (IMU). It is the prediction error covariance of the next moment obtained after inertial prediction.

[0073] S5. Integrity and Degradation Constraints: In the event of weak signals or loss of lock, non-GNSS information is used to update the navigation results with constraints.

[0074] Integrity monitoring is used to identify and eliminate abnormal observations. When the navigation source GNSS is in a degraded or unlocked state, but not completely failed, the inertial solution results are corrected by multi-source weighted fusion using barometric altitude and geomagnetic heading from S4. When integrity monitoring determines that the GNSS is completely failed, the following degradation mode constraint process is entered, which achieves continuity and stability of the navigation solution through three steps of coordinated operation: IMU calculation, barometric altitude and geomagnetic heading constraint, and covariance expansion.

[0075] The microcontroller unit (MCU) continuously performs residual-based outlier removal and RAIM / FDE integrity monitoring. When all GNSS is unavailable, external constraints are updated for inertial estimation using barometric altitude and geomagnetic heading, and the process noise covariance is adaptively adjusted to suppress long-term drift. In weak navigation mode, the external observation sampling rate and filter update frequency are reduced to decrease power consumption. Once the quality score recovers and remains above the threshold, the normal sampling and update strategy is resumed. The specific steps are as follows: 1) Anomaly detection removal In the EKF residual test, when the residual vector The standardized residual statistic (e.g., chi-square test) exceeds the threshold When the threshold is reached, the observation is considered abnormal. Based on the measurement vector dimension and significance level The corresponding chi-square distribution threshold is determined as follows:

[0076] That is, the standardized residual statistic is calculated as follows:

[0077] For the measurement matrix, To predict the covariance matrix, The corresponding measurement noise covariance matrix, For the first The residual vector of each measurement source.

[0078] When the model and noise assumptions are correct The observation vector dimension is and significance level Determine the threshold :

[0079] when If so, the measurement is deemed abnormal.

[0080] The dimension of the observation vector is The significance level is The chi-square distribution threshold; when the standardized residual statistic exceeds a confidence level of 1- If the chi-square threshold is reached, the measurement is considered abnormal and is removed (weighting is set). This avoids outliers from dragging down the fused solution and updates the fused input.

[0081] In one embodiment, when the observation vector dimension And the significance level is At that time, the anomaly detection threshold is =7.815; when the observation dimension is (Barometric altitude or geomagnetic heading) and significance level is At that time, the threshold is =3.841. The above threshold value is derived from the chi-square distribution table, which is a commonly used method in the field of navigation system consistency verification and fault detection.

[0082] 2) Source consistency check and integrity evaluation The multi-source consistency check refers to cross-comparing the solution results of different GNSS cores, judging whether the differences between sources are abnormal based on the covariance adaptive threshold (chi-square test), and triggering cross-validation and reducing the weight of the source when the difference exceeds the limit.

[0083] The overall integrity index calculation refers to calculating the navigation system protection level PL based on the filter covariance matrix and comparing it with the alarm limit AL to determine the navigation integrity. If the navigation system protection level PL exceeds the limit, a navigation unusable flag is output.

[0084] A. Source consistency check: The results of the aforementioned multiple multi-constellation GNSS cores (each GNSS core has an independent RF front-end and computing capabilities, and outputs its own navigation results) , , Perform a consistency comparison: If this occurs, cross-validation is triggered, reducing the weight of sources of inconsistency.

[0085] This is represented as the inter-source location difference threshold, used to determine the consistency of calculation results from different GNSS cores. This threshold is preferentially determined based on the chi-square distribution method. (Determined) The method originates from the chi-square test, which is calculated based on the statistical characteristics of the output covariance matrix of each GNSS core. It is a readily available statistical test method. The comparison is performed between any two GNSS cores with normal output, with the one having higher signal quality being preferred as the reference source. B. Overall Integrity Index (PL Protection Level)

[0086] in, For integrity factors (related to failure probability); if If so, output a navigation solution unavailable flag. Provided by business security requirements (functional safety / scenario requirements). This is an indicator of the overall integrity of the navigation system.

[0087] 3) Degradation Mode Constraints and Constraint Updates: In the degradation mode where GNSS is completely ineffective, the navigation system achieves continuity and stability of the navigation solution through three coordinated steps: IMU calculation, barometric altitude and geomagnetic heading constraints, and covariance expansion. Barometric and geomagnetic constraints suppress altitude and heading drift, respectively, while the covariance expansion mechanism reflects the trend of accuracy decline, ensuring the stability and reliability of the filtering process.

[0088] A. GNSS Complete Failure Scenario: After the GNSS signal fails, the navigation system cannot obtain external position observations. Therefore, the state update relies solely on the inertial calculation results of the IMU, while using barometric altitude and geomagnetic heading as constraint measurements for compensation and correction. ① Inertial prediction stage Based on the triaxial acceleration and angular velocity data output by the IMU, the MCU updates the navigation system state using inertial navigation equations to achieve continuous velocity, attitude, and position calculations. This process provides short-term continuity but is prone to drift.

[0089]

[0090] The predicted state value is the value obtained from the state at the previous time step. and control input Predict the current moment The state; ② Barometric Altitude Constraints Altitude measurement information provided by barometric pressure sensors With state prediction height By comparison, the pressure altitude consistency constraint is obtained: ; Barometric altitude constraint represents the maximum permissible difference between the measured barometric pressure and the predicted altitude. Barometric altitude constraint can effectively limit vertical drift and ensure the altitude stability of the navigation system.

[0091] ③ Geomagnetic heading constraints Navigation vector measurement information obtained using geomagnetic sensors Heading angle calculated from IMU Perform a consistency comparison: , Yaw angle consistency threshold indicates the permissible heading difference.

[0092] B. Covariance Inflation Strategy: In the degradation mode of GNSS signal failure, the observability of the navigation system will be significantly reduced. Therefore, to avoid the filter generating overly optimistic state estimates in the absence of external observations, the navigation system needs to use the process covariance matrix... Inflation is used to increase the conservatism of the state estimate, and the specific method is as follows:

[0093] The expansion coefficient (in one embodiment, the value ranges from 3 to 10) will take a larger value when the navigation system enters the degradation mode to reflect the degradation of the navigation system's accuracy.

[0094] 4) Soft constraints and state smoothing: Set rate of change constraints for each measurement source. This avoids position, velocity, and heading jumps caused by measurement rejection or degradation switching.

[0095] 5) Fusion result output and status feedback Output integrity flag: When the integrity flag is normal, output , When the integrity flag is abnormal or degraded, output the state estimate and the flag of the abnormal or degraded mode; If a source measurement is removed, S4 is notified to update the weights; or, if the system enters a degradation mode, the microcontroller unit (MCU) instructs the S4 module to use simplified measurement equations (referring to the sub-models of the GNSS, barometric, and geomagnetic measurement models established in step S4 under degradation mode, retaining only the measurement equations corresponding to currently available measurement sources and removing failed measurement sources) for state updates. The simplified measurement equations retain only valid measurement sources (such as barometric and geomagnetic constraints) and remove failed GNSS or other measurement sources. This ensures that the navigation system can maintain the continuity and stability of the navigation solution even when some navigation sources fail.

[0096] In one embodiment, the navigation system of the present invention can be widely used in drones, autonomous vehicles, precision surveying equipment, and high-reliability ship navigation systems. In complex environments with severe multipath effects or where GNSS signals are easily blocked (such as urban canyons, forest areas, and indoor-outdoor switching scenarios), a robust estimation of the navigation solution is achieved through multi-GNSS core parallel operation and a high-precision time synchronization mechanism. (Multi-GNSS core parallel operation: using wafer-level heterogeneous integration packaging, multiple multi-constellation GNSS cores and navigation processing cores are integrated in the same package in a parallel or partially stacked manner. Each GNSS core has multi-constellation receiving capability (supporting at least GPS, BeiDou, and GLONASS) and is equipped with an independent RF front-end. High-precision time synchronization mechanism: a multi-RF channel isolation structure and a unified clock / timestamp synchronization circuit are set within the package to achieve parallel sampling and time consistency of different GNSS receiving links. The unified clock / timestamp synchronization circuit uses the GNSS reference time to synchronize the IMU, geomagnetism, barometric pressure, and each GNSS receiving link, ensuring that the cross-sensor time alignment error is preferably no greater than 1 ms.) Combined with multi-source fusion in step S4, robust estimation of the navigation solution is achieved. This design can effectively reduce the risk of positioning interruption and solution divergence, thereby significantly improving the continuity, stability and integrity of navigation results.

[0097] At the packaging level, it achieves high integration and RF isolation, as well as clock unification. At the algorithm level, it implements a systematic solution of multi-source quality assessment, intelligent master-slave switching, smooth transition, and robust fusion, enabling the device to maintain continuous and highly stable navigation output in complex scenarios while miniaturizing and reducing power consumption.

[0098] Navigation Source: refers to multiple GNSS cores and their independent receiving links, mainly used for quality assessment, primary / backup switching and smooth transition in steps S1 to S3.

[0099] Sensor Source: refers to non-GNSS observation devices such as inertial measurement units (IMU), geomagnetic sensors, and barometric pressure sensors.

[0100] Measurement Source: Primarily used in the fusion update phase (steps S4-S5), this refers to external observation equipment such as GNSS, barometric pressure sensors, and geomagnetic sensors. It does not include an inertial measurement unit (IMU). Its output enters the measurement equations of the extended Kalman filter to correct inertial estimation errors. The acceleration and angular velocity output by the IMU, used as inputs to the state prediction model, are not considered part of the measurement source.

[0101] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined in this invention may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-source navigation signal switching and fusion method, characterized in that, The method comprises the following steps: quality assessment is performed on GNSS signals from different navigation sources, and a current main navigation source is determined according to the assessment result; when the quality score of the main navigation source is lower than a preset threshold or GNSS observation information meets a preset switching condition, switching of the main navigation source and a backup navigation source is performed, and smoothing processing is performed on old navigation solution and new navigation solution during the switching process; measurement data of the navigation source GNSS, an inertial measurement unit, a geomagnetic sensor and a barometric altitude sensor are fused and state estimation is performed through a filtering algorithm, and navigation results are output.

2. The method of claim 1, wherein, The quality assessment comprises calculating quality scores of the navigation sources and selecting the navigation source with the highest quality score as the current main navigation source, wherein the quality score of the navigation source is obtained based on GNSS observation information and through adaptive adjustment of a fusion weight.

3. The method of claim 1, wherein, The preset switching condition met by the GNSS observation information at least comprises one of the following: carrier signal-to-noise ratio is lower than a preset carrier lower limit; number of visible satellites is lower than a preset number value; position dilution of precision (PDOP) exceeds a preset precision threshold; carrier lock state is interrupted; cycle slip rate exceeds a preset cycle slip rate threshold.

4. The method of claim 1, wherein, The smoothing processing comprises: real-time calculation of position change, speed change and heading angle change of the old and new navigation solutions; when any change exceeds a preset soft constraint threshold, a first-order lag filter or spline interpolation is used to realize gradual transition; when the change is within the threshold range, the new main navigation source is directly switched to; a smoothing time constant is adaptively adjusted according to the quality difference between the main and backup sources and a carrier maneuverability index to control the gradual transition speed of the switching process.

5. The method of claim 1, wherein, The multi-source data fusion comprises: definition of a state vector of the navigation system and establishment of a state prediction equation; dynamic adjustment of a measurement noise covariance matrix according to real-time quality scores of each measurement source to realize adaptive weighting of GNSS position and speed measurement, barometric altitude measurement and geomagnetic heading measurement of the navigation source; obtaining the navigation state after multi-source fusion through filtering update.

6. The method of claim 5, wherein, The measurement model of each measurement source comprises: Measurement model for GNSS position of primary navigation source: , is the GNSS position noise of primary navigation source, which is assumed to be zero-mean Gaussian white noise, , measurement noise covariance matrix of GNSS of primary navigation source Adaptive adjustment according to the quality score of GNSS of primary navigation source; The measurement model of the barometric altitude sensor is: , is the barometric altitude noise, which is assumed to be zero-mean Gaussian white noise, , is the barometric altitude measurement noise covariance matrix, , is the reference noise variance of the barometric altitude measurement, is the real-time quality score of the barometric sensor, is a tiny constant to prevent the denominator from being zero; The measurement model of the geomagnetic sensor is , is the yaw angle in the attitude ; is the geomagnetic heading angle noise, which is assumed to be zero-mean Gaussian white noise, , is the geomagnetic measurement noise covariance matrix, , is the reference noise variance of the geomagnetic heading measurement, is the real-time quality score of the geomagnetic sensor, is a tiny constant to prevent the denominator from being zero.

7. The method of claim 1-6, wherein, The method further comprises the steps of: performing abnormal measurement elimination and integrity monitoring to identify and eliminate abnormal measurement; when the navigation source GNSS is in degradation or lock loss but not completely failed, using barometric altitude and geomagnetic heading as external measurement to correct the inertial navigation result; when the navigation source GNSS is completely failed, the inertial measurement unit, barometric altitude and geomagnetic heading are used for constraint and covariance inflation to realize continuous and stable navigation results.

8. A navigation system based on multi-navigation module and wafer level integration, characterized in that, Wafer-level heterogeneous integration packaging is used to integrate multiple multi-constellation GNSS chiplets and navigation processing chiplets in the same package in a side-by-side arrangement or local stacking manner; the navigation processing chiplet comprises a micro control unit, an inertial measurement unit, a geomagnetic sensor and a barometric sensor; the micro control unit is configured with a control program for realizing the method of any one of claims 1-7.

9. The navigation system based on multi-navigation module and wafer level integration of claim 8, wherein, A multi-radio channel isolation structure and a unified clock / time stamp synchronization circuit are arranged in the package, and the heterogeneous chiplets are interconnected through a shared data bus and a power supply network in the package, and a package-level electromagnetic shielding structure is arranged to reduce radio frequency interference and interconnection parasitic effects.

10. The navigation system based on multi-navigation module and wafer level integration of claim 8, wherein, The wafer-level heterogeneous integration at least uses one or more processes of through-silicon via, micro-bump bonding and re-distribution layer to realize vertical interconnection and high-density wiring of the heterogeneous chiplets; and / or, the GNSS chiplet and the navigation processing chiplet are interconnected laterally through RDL or vertically through TSV.

Images