A GNSS / LEO / MIMU observation lock loss bridging and hierarchical re-convergence method based on low earth orbit satellite signal assistance

Abstract

The application discloses a GNSS / LEO / MIMU observation lock loss bridging and hierarchical re-convergence method based on low-orbit satellite signal assistance, which acquires GNSS observation data, low-orbit satellite LEO auxiliary observation data and MIMU inertial measurement data, and establishes a GNSS / LEO / MIMU tight combination navigation error state model; candidate observation is constructed in view of information gain, shielding reliability and prediction consistency joint evaluation index, so that observation screening and weight adjustment are realized; according to the number of effective observations, the number of LEO auxiliary observations, information gain and continuous lock loss time, running mode discrimination is carried out, and under the condition of GNSS short-time lock loss, the LEO auxiliary bridging mode is entered; in the bridging process, Schmidt sub-state frozen filtering is adopted, only the active state is updated and the frozen state is propagated; when the external observation is recovered, covariance inflation is carried out according to the lock loss time, and the frozen state is updated and recovered hierarchically in combination with the sliding window consistency criterion. The application can improve the continuous positioning capability and recovery stability of the GNSS / LEO / MIMU integrated navigation system under complex environment.

Description

Technical Field

[0001] This invention belongs to the field of satellite navigation and integrated navigation technology, specifically relating to a GNSS / LEO / MIMU observation loss-of-lock bridging and hierarchical reconvergence method based on low-orbit satellite-aided information, which is applicable to continuous positioning in urban canyons, complex obstruction, and GNSS observation degradation environments. Background Technology

[0002] GNSS possesses all-weather, high-precision positioning capabilities and is a crucial information source for satellite navigation and integrated navigation systems. However, it is susceptible to obstruction, interference, and multipath effects in urban canyons, densely built-up areas, and complex electromagnetic environments, leading to a reduction in the number of observations or even short-term loss of lock, thus degrading the accuracy of GNSS / LEO / MIMU integrated navigation. Existing technologies have the following shortcomings: 1. Observation selection only considers reliability, not information contribution; 2. LEO-assisted state management is lacking during lock-out; 3. State abrupt changes are prone to occur during the recovery phase.

[0003] LEO (Low Orbit Satellite) has advantages such as low orbit, strong signal, and the ability to provide additional observation constraints, making it a valuable source of auxiliary information during GNSS lock-up. However, existing methods are mostly simple fusions, lacking a unified state management approach for the entire process of "GNSS lock-up—LEO-assisted bridging—recovery and reconvergence." While some methods introduce the concept of state freezing during lock-up, they fail to define the applicable duration of the freezing strategy. Furthermore, directly restarting all state estimation after GNSS recovery may cause filter oscillations or even divergence due to insufficient expansion of the prior covariance of the frozen state. In complex environments, GNSS signals may experience continuous obstruction or short-term lock-up. In such cases, traditional GNSS / INS fusion methods typically rely on inertial propagation to maintain navigation state. However, due to the accumulation of inertial device errors over time, prolonged lock-up can lead to a rapid increase in positioning errors. Simultaneously, if full-state filtering updates are performed directly after external observation recovery, weakly observable states may be affected by abnormal observations, resulting in significant fluctuations in state estimation. Therefore, a method is needed that can utilize LEO satellite-assisted observations to maintain state constraints during lock-up and achieve stable and asymptotic reconvergence after observation recovery. Summary of the Invention

[0004] Objective: To address the aforementioned problems, this invention provides a method and system for bridging and hierarchical reconvergence of GNSS / LEO / MIMU observations based on low-Earth orbit (LEO) satellite signals. This invention achieves candidate observation screening and weight adjustment by constructing a joint observation evaluation model that considers information gain, occlusion reliability, and prediction consistency; it achieves switching between normal mode, degradation mode, bridging mode, and pure inertial propagation mode through an outer-layer operation mode discrimination mechanism; in bridging mode, it employs a Schmidt sub-state freezing strategy, updating only reliably estimable active states; after observation recovery, it drives the hierarchical recovery and updating of frozen states through covariance inflation and statistical consistency criteria.

[0005] Technical solution: The present invention adopts the following technical solution, including the following steps.

[0006] (1) Acquire GNSS observation data, LEO-assisted observation data, and MIMU inertial measurement data; establish a GNSS / LEO / MIMU tightly coupled navigation error state model; and use the error state extended Kalman filter for state prediction. The system error state is represented as:

[0007] ;

[0008] in, These represent position error, velocity error, and attitude error, respectively. These are the zero bias of the gyroscope and the zero bias of the accelerometer, respectively. These are receiver clock bias and clock drift, respectively. The deviation between GNSS and LEO systems; This is due to residual tropospheric wet delay;

[0009] (2) For GNSS and LEO candidate observations, a joint evaluation model is established, considering information gain, occlusion reliability, and prediction consistency. This model is used to select effective external constraint information in the case of GNSS observation degradation or loss of lock, achieving hard screening and soft weighting of observations without changing the underlying compact combination filter structure. This evaluation mechanism adopts a unified evaluation framework for GNSS and LEO observations, but in bridging mode, it is mainly used to screen and adjust LEO auxiliary observations participating in the update. Based on the key error sub-state covariance after dimensionless processing of the reference scale matrix, for the i-th candidate observation, the information gain and its normalized form are:

[0010] ;

[0011] in, To incorporate the covariance of the key error sub-states prior to this observation, This represents the critical error substate posterior covariance calculated based on the single observation update after adding the i-th candidate observation. For the observation dimension; The elevation and azimuth angles of the observation line are given. The critical elevation angle given for the street valley shading boundary model; Let the observed innovation vector and innovation covariance be given. This serves as a consistency threshold. Based on this, a joint evaluation index is constructed, and hard screening and soft weighting are implemented:

[0012] ;

[0013] when ,or ,or When this happens, the candidate observation is removed from the valid observation set; The original observation noise covariance, These are the joint score and the occlusion reliability threshold, respectively.

[0014] (3) Let the effective observation set at the current epoch be . The quantity is The number of opportunistic LEO observations is denoted as ;set up Corresponding to the updated key error sub-state covariance Before the update Define the overall information gain of the key error substate (with the same construction form as the single observation information gain in step 2, the difference being that it applies to the entire substate here). Perform a hypothetical joint update):

[0015] ( When it is an empty set );

[0016] =Normal / Degenerate / Bridging / Pure Inertial Propagation, corresponding in that order;

[0017] ① and ;

[0018] ② ;

[0019] ③ The current GNSS observations are lost or there are insufficient effective GNSS observations, and ;

[0020] ④ Otherwise

[0021] Corresponding update strategies: Normal mode features robust full-state update; Degenerate mode still updates all states, but since the scoring reduction in step 2 has already been completed on the front end, the update intensity is naturally weakened, requiring no additional mechanism; Bridge mode proceeds to the Schmidt freeze update section below; Pure inertial propagation mode only propagates state. Bridge mode is only applicable to... No more than Short-term scenarios: when Exceed ,or When this occurs, the system exits bridging and switches to pure inertial propagation mode, and enters the inner recovery state machine described below after recovery. In bridging mode, when the number of effective GNSS observations is insufficient, opportunistic LEO-assisted observations provide additional external constraints to the integrated navigation system to reduce error accumulation caused by pure inertial propagation.

[0022] To avoid mode jitter near the threshold, a hysteresis switching strategy is adopted: rapid degradation and slow upgrade: the switch is made immediately when degrading from a high level to a low level; the upgrade can only be performed after the target conditions have been met continuously for a preset number of epochs when recovering from a low level to a high level.

[0023] (4) In bridging mode, the error state is divided into active state and frozen state, using the valid observations retained after the screening in step 2 (including opportunistic LEO observations and possibly residual valid GNSS observations) as input:

[0024] ;

[0025] The observation matrix and prior covariance are expressed as follows:

[0026] ;

[0027] The information covariance under bridging mode is:

[0028] ;

[0029] The active state Schmidt gain is:

[0030] ;

[0031] Status updated to:

[0032] ;

[0033] in, This represents the observation residual vector that actually participates in the update in bridging mode. Its sources include opportunistic LEO observations and potentially residual valid GNSS observations. Since the frozen state does not participate in the measurement feedback update during the lockout period, its estimation uncertainty increases over time. Therefore, during the recovery phase, the frozen state covariance is adjusted so that the recovered observation update can fully reflect the uncertainty accumulated during the lockout period.

[0034] Covariance is updated in blocks using the Schmidt form:

[0035] IDT block-based update:

[0036] ;

[0037] In this way, the frozen state does not participate in the current measurement feedback update, but its statistical correlation with the active state is still preserved through cross-covariance.

[0038] (5) Once external observations are restored, the system initiates a recovery state machine independent of the outer operating mode. The recovery phase first considers the continuous lockout time prior to recovery. Adjustments are made for the uncertainty of the frozen state. The frozen state covariance inflation factor is defined as follows:

[0039] ;

[0040] The inflated prior covariance is:

[0041] ;

[0042] in, The coefficient of thermal expansion is 1 / 3. This is the upper limit of expansion. These are identity matrices representing the dimensions of the active and frozen states, respectively.

[0043] Subsequently, based on the statistical characteristics observed during the recovery phase, the sliding window average normalized innovation square (ANIS) was adopted as the state release criterion:

[0044] ;

[0045] in, To restore the length of the sliding window for judgment, These represent the observation residual vector, information covariance, and observation dimension for the j-th valid update epoch. Based on the number of observations, overall information gain, and ANIS statistical consistency, the frozen state is restored step-by-step in order of increasing difficulty. The reconvergence process is then divided into three sequentially executed stages:

[0046] Phase I

[0047] Status of open for updates:

[0048] Entry conditions: External observations become available again and

[0049] This phase primarily involves restoring the position, velocity, and receiver time-related states.

[0050] Phase II

[0051] Status of open for updates:

[0052] Entry requirements:

[0053] The preset epoch number is continuously satisfied.

[0054] Phase III

[0055] Status of open for updates:

[0056] Entry requirements:

[0057] The preset epoch number is continuously satisfied.

[0058] After completing Phase III, the system regains its full state update capability. If external observations deteriorate again during the recovery process, or if the ANIS index deviates from the corresponding range, the current recovery phase is terminated, and the system re-enters the operational mode for discrimination. Since the observability and convergence speed of different states vary, this method sequentially opens updates according to the difficulty of state recovery, from low to high. Position, velocity, and time-related states are directly constrained by external observations and are therefore recovered first; attitude errors and inertial device biases are significantly affected by system dynamic errors and are recovered subsequently; slowly varying parameters have stronger coupling and are recovered last.

[0059] Beneficial effects: Compared with the prior art, the present invention has the following beneficial effects:

[0060] (1) This invention establishes a joint evaluation model for information gain, occlusion reliability and prediction consistency for GNSS / LEO candidate observations, realizes hard screening and soft weighting of observations, and improves the rationality of observation utilization in complex environments.

[0061] (2) The present invention establishes a multi-mode operation discrimination mechanism based on the number of effective observations, information gain and lockout time, and sets a maximum duration constraint for bridging mode to avoid the long-term maintenance of the frozen state.

[0062] (3) The present invention adopts the Schmidt substate freezing strategy under the condition of short-term GNSS lockout, and only performs measurement feedback on the active states that can be reliably estimated, thereby reducing the risk of weak observable states being affected by abnormal observations.

[0063] (4) In the observation recovery stage, the present invention constructs a frozen state covariance expansion mechanism based on the unlocking time, and releases the frozen state step by step in combination with the ANIS statistical consistency index to improve the stability of the recovery process.

[0064] (5) This invention combines the outer layer operation mode discrimination with the inner layer recovery state machine to achieve unified state management of the entire process of GNSS loss of lock, LEO assisted bridging and recovery reconvergence. Attached Figure Description

[0065] Figure 1 This is a flowchart illustrating an embodiment of the present invention. Detailed Implementation

[0066] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. For example... Figure 1 As shown, the GNSS / MIMU fusion positioning enhancement method based on low-Earth orbit satellite signals of the present invention includes the following steps:

[0067] Step 1: Acquire GNSS observation data, LEO auxiliary observation data, and MIMU output inertial measurement data. Establish a compactly integrated navigation error state model based on the GNSS, LEO, and MIMU observation information:

[0068] ;

[0069] Error state extended Kalman filtering is used for state prediction and covariance propagation, providing predicted states for subsequent observation evaluation and mode switching.

[0070] Step 2: For each epoch of GNSS and LEO candidate observations, the information gain, occlusion reliability factor, and prediction consistency factor are calculated according to the following formulas:

[0071] ;

[0072] ;

[0073] ;

[0074] Further construct joint evaluation indicators .

[0075] Step 3: Based on the valid observation set Number of valid observations Number of opportunistic LEO observations and the overall information gain of key error substates Combined with the duration of continuous lock loss Determine the current operating mode.

[0076] For example, suppose that statistics from a certain epoch yielded... Current duration of continuous lockout Preset parameters .because The conditions for normal mode are not met; due to the insufficient number of currently effective GNSS observations, the information gain of the key error substate is below the degradation threshold, while there are opportunistic LEO-assisted observations that meet the conditions. and According to the priority order described in this invention, the system determines that it is in bridging mode and proceeds to Example 4. If the lock loss continues thereafter, when Growth exceeded Or the number of opportunistic LEO observations drops to The system will no longer maintain the same frozen bridging structure, automatically exit the bridging mode and switch to the pure inertial propagation mode. After the external observation is restored, it will uniformly enter the inner recovery state machine described in Example 5.

[0077] Step 4: Divide the error state into active state and frozen state:

[0078] ;

[0079] The observation matrix and prior covariance are expressed as follows:

[0080] ;

[0081] The information covariance under bridging mode is:

[0082] ;

[0083] The active state Schmidt gain is:

[0084] ;

[0085] Status updated to:

[0086] ;

[0087] Simultaneously, the covariance matrix is ​​updated according to the Schmidt form. This process maintains the statistical correlation between the frozen and active states. Since the number of external observations decreases during the unlocking period, the observability of some states declines. If full-state feedback updates continue, weakly observable states are easily affected by unreliable observations. Therefore, this method only updates measurements for the currently observable states and freezes weakly observable states to maintain the stability of state estimation.

[0088] Step 5: Once GNSS or other external observations are restored, trigger the recovery state machine. First, calculate based on the continuous loss of lock time:

[0089] ;

[0090] And construct the frozen state covariance expansion matrix:

[0091] ;

[0092] The inflated prior covariance is:

[0093] ;

[0094] Then, the sliding window statistics were calculated:

[0095] ;

[0096] Execute sequentially based on observation quality:

[0097] (1) Restore activity status update;

[0098] (2) Recover attitude error and update gyroscope zero bias;

[0099] (3) Restore accelerometer zero bias, inter-system bias and residual tropospheric wet delay update.

[0100] Once the recovery phase meets the corresponding information gain and statistical consistency conditions, the set of states participating in the update is gradually expanded. If the observation quality deteriorates during the recovery process, the system returns to the operating mode determination phase to redetermine the system state.

[0101] For example, let the length of the recovery decision sliding window be... Phase II corresponds to the consensus interval Phase III corresponds to a narrower range. If, after external observations are restored, the calculated ANISK = 1.05 for 10 consecutive epochs, and... If both conditions are met, the system will unfreeze from Phase I and enter Phase II, restoring the filter update for attitude error and gyroscope zero bias state; if subsequently... Continue to fall If the system is further unfrozen, it will enter Phase III, restoring the filtering updates for the remaining frozen states, such as accelerometer zero bias, inter-system deviation, and residual tropospheric wet delay, and will be reintegrated into the online discrimination range between normal mode and degradation mode.

[0102] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for GNSS / LEO / MIMU observation lockout bridging and hierarchical reconvergence based on low-Earth orbit satellite signal assistance, characterized in that, Includes the following steps: S1: Acquire GNSS observation data, LEO-assisted observation data, and MIMU inertial measurement data; establish a GNSS / LEO / MIMU tightly coupled navigation error state model; and use error state extended Kalman filtering for state prediction and covariance propagation; its state vector is: ； in, These represent position error, velocity error, and attitude error, respectively. These are the zero bias of the gyroscope and the zero bias of the accelerometer, respectively. These are receiver clock bias and clock drift, respectively. The deviation between GNSS and LEO systems; This is due to residual tropospheric wet delay; S2: For GNSS and LEO candidate observations, an information gain index is constructed based on the change of key error sub-state covariance, and a joint evaluation model is established by combining the observation reliability factor and the prediction consistency factor to screen candidate observations and adjust their weights. S3: The system operation mode is determined based on the number of effective observations, the number of opportunistic LEO observations, the information gain of key error substates, and the continuous lockout time. The system operation modes include normal mode, degradation mode, bridging mode, and pure inertial propagation mode, and the switching between different modes is achieved through a hysteresis switching mechanism. S4: When the system enters the bridging mode, the error state is divided into active state and frozen state. The Schmidt substate freezing filtering method is adopted. Measurement feedback updates are only performed on the active state. The frozen state does not participate in measurement feedback but maintains state propagation and covariance propagation. S5: Once external observations are restored, adjust the uncertainty of the frozen state according to the continuous loss of lock, so as to gradually restore the frozen state. S6: Based on the number of observations, information level, and statistical consistency criteria of the sliding window during the recovery phase, the frozen state is opened and updated in a preset order to achieve hierarchical reconvergence.

2. The method for GNSS / LEO / MIMU observation lock-off bridging and hierarchical reconvergence based on low-Earth orbit satellite signals according to claim 1, characterized in that, The joint evaluation model constructs an information gain index based on the change in the covariance of key error sub-states. For the i-th candidate observation, its information gain is expressed as: ； in, To incorporate the covariance of the key error sub-states prior to this observation, This represents the critical error substate posterior covariance calculated based on the single observation update after adding the i-th candidate observation.

3. The method for GNSS / LEO / MIMU observation lock-off bridging and hierarchical reconvergence based on low-Earth orbit satellite signals according to claim 2, characterized in that, The information gain is normalized, and its normalization form is as follows: ； Based on the normalized information gain, observation reliability, and prediction consistency information, candidate observations are screened or their weights are adjusted.

4. The method for GNSS / LEO / MIMU observation lockout bridging and hierarchical reconvergence based on low-Earth orbit satellite signal assistance as described in claim 1, characterized in that, In the bridging mode, the error state is divided into an active state and a frozen state: ； in, Indicates the activity status. This indicates a frozen state; frozen states do not participate in current measurement feedback updates, but maintain state propagation and covariance correlation with active states.

5. The method for GNSS / LEO / MIMU observation loss-of-lock bridging and hierarchical reconvergence based on low-Earth orbit satellite signals according to claim 4, characterized in that, The hierarchical reconvergence uses a sliding window statistical consistency criterion to determine the release conditions of the frozen state; when the observation consistency within the recovery window meets the preset conditions, it enters the next stage of state recovery.

6. The method for GNSS / LEO / MIMU observation lock-off bridging and hierarchical reconvergence based on low-Earth orbit satellite signal assistance as described in claim 1, characterized in that, Once external observations are restored, an uncertainty inflation factor for the frozen state is constructed based on the duration of continuous loss of lock before restoration: ； in, The coefficient of thermal expansion is 1 / 3. The upper limit of the expansion is given; the expansion matrix of the frozen state covariance is constructed based on the expansion factor, and the prior covariance of the frozen state is adjusted.

7. The method for GNSS / LEO / MIMU observation lock-off bridging and hierarchical reconvergence based on low-Earth orbit satellite signals according to claim 1, characterized in that, During the hierarchical reconvergence process, a sliding window statistical consistency index is used to determine the conditions for releasing the frozen state; when the observation consistency in the recovery phase meets the preset conditions, the next phase of state recovery begins.

8. The method for GNSS / LEO / MIMU observation lock-off bridging and hierarchical reconvergence based on low-Earth orbit satellite signal assistance as described in claim 1, characterized in that, The graded reconvergence includes: the first stage, restoring the position, velocity, and time-related states; the second stage, restoring the attitude error and inertial device bias states; and the third stage, restoring other frozen states. Each stage is switched according to the observation consistency and state convergence during the restoration process.

Images