Method for multipath mitigation of carrier loop based on improved sage-husa adaptive kalman filter

Abstract

The application discloses a carrier loop multipath suppression method based on improved Sage-Husa adaptive Kalman filtering, which comprises the following steps: signal correlation processing, early, positive and late correlation outputs and I / Q outputs are obtained; performing multipath detection, early and late correlation ratio analysis, phase lock quality evaluation and code phase deviation dynamic analysis are performed; calculating the basic measurement value, obtaining the preliminary code phase error and the carrier phase error; using carrier phase smoothing processing to enhance the measurement value, adjusting the noise covariance according to the multipath severity, performing adaptive Kalman filtering, including state prediction, gain calculation, state update, Sage-Husa adaptive noise covariance update and adaptive factor dynamic adjustment, storing data and preparing for the next cycle. Through the combination of multipath detection and adaptive filtering, the application effectively identifies the multipath interference and dynamically adjusts the system parameters, and significantly improves the phase tracking precision and reliability of the GNSS signal in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of GNSS satellite signal processing technology, and in particular to a carrier loop multipath suppression method based on an improved Sage-Husa adaptive Kalman filter. Background Technology

[0002] Global Navigation Satellite Systems (GNSS) play a crucial role in modern navigation and positioning applications. However, in real-world environments, GNSS signals are often severely affected by multipath effects. This effect occurs when signals are reflected by obstacles such as buildings and terrain, causing the receiver to receive both direct and reflected signals simultaneously. This can lead to phase tracking errors in the receiver, significantly reducing navigation and positioning accuracy.

[0003] During satellite signal reception, the signal received by the echo antenna, besides the detected target echo, is mixed with many other signals, such as satellite direct waves, various noises, and ground object reflection clutter. Furthermore, the power of the target signal is usually much lower than the power of these clutters. Therefore, it is necessary to process the echo signal to extract the target echo containing target information from the numerous signals. The carrier tracking loop is one of the most critical components of a GNSS receiver, and its performance directly determines the accuracy of phase measurement. Traditional carrier tracking techniques mainly employ phase-locked loop (PLL) structures. Although simple to implement, they are prone to locking errors or even loss of lock under multipath interference. To improve the robustness of the carrier loop in multipath environments, tracking methods based on estimation theory have been introduced, among which Kalman filtering technology has received widespread attention due to its excellent state estimation capabilities. However, the performance of the standard Kalman filter largely depends on the accurate modeling of the system and the statistical characteristics of the measurement noise. In actual multipath environments, noise characteristics often exhibit non-static and non-Gaussian features, significantly differing from the white noise model assumed by the standard Kalman filter. This model mismatch can lead to filter performance degradation and even divergence.

[0004] Furthermore, most existing multipath suppression techniques focus on signal correlation processing, such as multiple correlator (MCR) and narrow correlator (NCR) techniques. While these techniques can mitigate the effects of multipath interference to some extent, multipath interference is not static. Traditional structures (NCR and MCR) rely on fixed time delays and cannot adapt to changes in multipath delays in real time. Additionally, these techniques typically increase hardware complexity, hindering receiver miniaturization and low-power design.

[0005] In contrast, implementing multipath suppression at the tracking loop level offers greater flexibility and lower implementation costs. By improving the tracking algorithm, especially the adaptive filtering algorithm, the level of trust in the observation data can be dynamically adjusted without increasing hardware complexity, effectively improving the system's multipath resistance. Summary of the Invention

[0006] The problem to be solved by this invention is to provide a carrier loop multipath suppression method based on an improved Sage-Husa adaptive Kalman filter. By establishing a multi-layer multipath detection mechanism, dynamically adjusting the noise covariance matrix, and improving the adaptive factor update strategy, this method achieves high-precision phase tracking of GNSS signals in complex multipath environments, providing a new technical solution for improving the positioning accuracy of GNSS receivers in harsh environments.

[0007] This invention adopts the following technical solution: a carrier loop multipath suppression method based on improved Sage-Husa adaptive Kalman filtering, comprising the following steps:

[0008] Step 1, Signal Correlation Processing: Generate three sets of local reference signals with early, positive, and late time offsets, perform correlation operations with the input satellite signals respectively, and extract the time delay error using the in-phase and quadrature outputs to obtain the correlation results of the early, positive, and late channels;

[0009] Step 2, Multipath Detection: Analyze the ratio of the correlation results of the early and late paths to determine the phase lock quality and code phase tracking fluctuations. Use a multipath detection system to detect whether multipath effects exist, identify and assess the existence and severity of multipath interference.

[0010] Step 3, Basic Measurement Calculation: Calculate the code phase error using the correlation outputs of the early and late paths, and calculate the preliminary carrier phase tracking error using the phase relationship between the in-phase and quadrature components of the positive branch;

[0011] Step 4, Measurement Enhancement: The original carrier phase measurement value is smoothed and automatically adjusted according to the severity of multipath, reducing the phase error caused by noise and multipath effects before entering the filter;

[0012] Step 5, Dynamic adjustment of noise covariance: Based on the severity of the detected multipath, dynamically adjust the Kalman filter's prediction of the measurement noise covariance and process noise covariance;

[0013] Step 6: Adaptive Kalman Filtering: Using Kalman filtering combined with the Sage-Husa algorithm, noise parameters are dynamically adjusted to estimate the carrier phase, frequency, and code phase state of the signal; the fixed attenuation factor in the adaptive Kalman filter is replaced with an adaptive factor that is dynamically adjusted according to the severity of multipath, so as to accurately track the carrier phase.

[0014] Step 7, Tracking Parameter Update: Update the control parameters of the receiver's internal tracking loop using the estimation results of the Kalman filter;

[0015] Step 8, Data Storage and Iteration: Save the processing results and status of the current cycle to prepare for the next round of tracking calculation, and realize carrier loop multipath suppression of Kalman filtering.

[0016] As a preferred embodiment, in step 1, the signal correlation processing includes:

[0017] Step 1.1: Mix the received satellite signal with two orthogonal local carrier waves. and Multiply the signal, remove the carrier from the received signal, and convert the signal back to baseband to obtain the baseband in-phase signal after carrier stripping. signals orthogonal to baseband ;

[0018] Step 1.2: The signal after carrier stripping and The signals are multiplied by the three local reference signals of morning, mid-day, and evening respectively, and each signal is despread to obtain the three baseband in-phase and quadrature signals of morning, mid-day, and evening respectively.

[0019] Step 1.3, within a predetermined time interval Within this process, the early, positive, and late baseband in-phase and quadrature signals after despreading are accumulated separately to obtain the three-channel correlation results, which are expressed as follows: , , , , , ;

[0020] Among them, subscript E , P , L These correspond to the three local reference signal code copies for morning, noon, and evening respectively. P The code represents a local code copy that is phase-aligned with the current best estimated received code. E code and L The codes represent the ratios respectively. P A local copy of the code that is advanced and delayed by a certain number of code chips; , for E Results of in-phase and orthogonal integrals of branches , for P Results of in-phase and orthogonal integrals of branches , for L Results of in-phase and orthogonal integrals of the branches.

[0021] As a preferred embodiment, in step 2, the multipath detection system employs three complementary methods to identify the presence and severity of multipath signals, including:

[0022] Step 2.1: Morning and Evening Asymmetric Analysis. The difference between the complex signal amplitudes output by the correlation results of the morning and evening paths is used as the basis for judging whether multipath interference exists.

[0023] Step 2.2, Code error variance monitoring, the variance of the code error is expressed as:

[0024] ;

[0025] in, The variance of the total tracking error is represented by the following: This represents the variance of the error caused by noise. This represents the error variance caused by multipath effects; when multipath exists, The variance increases, exceeding the variance level in a purely noisy environment;

[0026] Step 2.3, Phase-locking quality assessment, with the following quantitative indicators:

[0027] ;

[0028] in, It is a carrier tracking quality indicator used to reflect the locking status of the carrier loop.

[0029] Furthermore, the multipath effect is determined using a voting mechanism, which calculates the detection results of three indicators from three complementary methods. When at least two indicators are judged to be multipath, multipath is considered to exist, and the severity of multipath is calculated as supplementary information in the detection output.

[0030] As a preferred option, in step 3, the code phase error is initially calculated using the correlation outputs of the early and late paths. Output via the discriminator:

[0031] ;

[0032] in, and These represent the power of the morning and evening branches, respectively;

[0033] Using the phase relationship between the in-phase and quadrature components of the positive branch, the preliminary carrier phase tracking error is calculated. Output via the discriminator:

[0034] .

[0035] As a preferred option, in step 5, the measurement noise covariance matrix is ​​adjusted as follows:

[0036] Step 5.11: Define the initial measurement noise covariance matrix. :

[0037] ;

[0038] in, Indicates code phase noise, This represents carrier phase noise.

[0039] Step 5.12: Based on the multipath severity detected in step S2, define the code phase noise gain coefficient and the carrier phase noise gain coefficient respectively, and apply them to the matrix. Dynamically scale the main diagonal elements:

[0040] ;

[0041] ;

[0042] in, and These represent the code observation noise scaling factor and the carrier phase observation noise scaling factor, respectively. and Representation matrix The scaled value of the main diagonal element;

[0043] Step 5.13: Divide the scaled covariance matrix by the loop time to obtain the final measurement noise covariance. .

[0044] Furthermore, the process noise covariance matrix is ​​adjusted as follows:

[0045] Step 5.21: Define the initial process noise covariance matrix. ;

[0046] Step 5.22: Based on the multipath severity detected in step S2, define the code phase state gain coefficient, carrier phase state gain coefficient, frequency state gain coefficient, and frequency change rate state gain coefficient, respectively, and apply them to the matrix. The diagonal elements are dynamically scaled to obtain a scaled matrix. ;

[0047] Step 5.23, based on the scaled matrix Through the discretized form of the state transition matrix and time-related change matrix The discretized process noise covariance matrix at the current time is obtained. .

[0048] As a preferred option, in step 6, Kalman filtering combined with the Sage-Husa algorithm is used to dynamically adjust the noise parameters, combining the innovation and Kalman gain. Regarding the current moment process noise mean Covariance and the mean of observation noise Covariance Perform recursive estimation;

[0049] ;

[0050] ;

[0051] ;

[0052] ;

[0053] in, Indicates the previous moment The estimated value of the mean process noise, , Indicates the current time The previous moment State estimates, This is the state transition matrix;

[0054] Indicates the previous moment The estimated value of the process noise covariance matrix, It is the current moment. The posterior estimation error covariance matrix, It was the previous moment The posterior estimation error covariance matrix;

[0055] Indicates the receiver at the current time. The observation vector generated during loop update is obtained by the tracking loop at the current time. The correlator output is calculated by the discriminator and used to characterize the code alignment error observation and carrier phase alignment error observation of the current loop. It is the observation matrix;

[0056] Indicates the previous moment The estimated value of the mean measurement noise. This represents the prior estimate of the state vector. Indicates the previous moment The estimated value of the measurement noise covariance, It is the prior estimation error covariance; This is a fixed decay factor used to balance the impact of new and old data on the current estimate.

[0057] As a preferred option, in step 6, the fixed attenuation factor in the adaptive Kalman filter is... Replace with an adaptive factor that is dynamically adjusted based on the severity of multipath attacks. When multipath interference is detected, adjustments are made based on the severity of the multipath interference and the tracking time. Adjustments are made to the adaptive factor. The calculation is as follows:

[0058] ;

[0059] Where 0.95 is the maximum allowable weighting factor, used to avoid the influence of adaptive factors during the estimation of noise covariance. Excessive weighting can cause drastic fluctuations in the filter update weights;

[0060] The calculation process for the update rate is as follows:

[0061] ;

[0062] ;

[0063] in, The time decay factor, This represents the number of filtering iterations. Additional adjustment rate; Based on the update rate, Multipath severity;

[0064] In the initial stage, the more severe the multipath effect, The larger, The larger; in subsequent stages, for Adjustments are made to arrive at the final update rate. .

[0065] As a preferred embodiment, in step 7, the noise statistics are estimated using a Kalman filter. , Dynamically adjust the bandwidth of the loop filter:

[0066] An increased value indicates a poor observation environment with severe multipath propagation and a low signal-to-noise ratio, necessitating a reduction in loop bandwidth. To suppress noise;

[0067] Increasing the loop bandwidth indicates higher receiver dynamism, requiring a larger loop bandwidth. To quickly track signal changes.

[0068] Through adaptive algorithms, based on and Real-time calculation of the optimal noise equivalent bandwidth of the carrier loop and code loop :

[0069] ;

[0070] Among them, the function Used to dynamically calculate the optimal filter bandwidth or control parameters based on real-time estimates of system noise and measurement noise; the calculated... The coefficients are converted into carrier loop and code loop filter coefficients, which determine the loop's response speed and smoothness to discriminator errors, thus optimizing the adjustment process of the carrier frequency and code frequency.

[0071] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0072] 1. The method of this invention combines multi-layer multipath detection and adaptive filtering technology to dynamically map the multipath detection results to the Kalman filter parameters and implement the correlation processing and multipath detection functions in parallel. The improved Sage-Husa algorithm is used to optimize the subsequent noise covariance estimation and state update operations. Finally, the dynamic adaptive factor adjustment mechanism is used to quickly adapt to different multipath intensity environments.

[0073] 2. The method of the present invention adopts multi-level detection, adaptive parameter adjustment and parallel processing design to enhance the system's resistance to multipath interference, improve the phase tracking accuracy and reliability of GNSS receiver in complex multipath environments, and effectively reduce the tracking error caused by multipath. It can be dynamically adjusted according to the signal environment. Attached Figure Description

[0074] Figure 1 This is a flowchart of the carrier loop multipath suppression method of the present invention. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the application will be further described in detail below with reference to the accompanying drawings. The described embodiments are only a part of the embodiments involved in this invention. All non-innovative embodiments based on these embodiments by other researchers in the art are within the protection scope of this invention. Furthermore, the step numbers in the embodiments of this invention are only set for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0076] In one embodiment of the present invention, a carrier loop multipath suppression method based on an improved Sage-Husa adaptive Kalman filter is provided, the processing flow of which is as follows: Figure 1 As shown, it includes the following steps:

[0077] S1, Signal Correlation Processing: Performs correlation processing on the input GNSS signal with the early signals generated locally by the receiver. E ),just( P ),Night( L The mixing and integration of the three local code copies and the carrier outputs a series of raw in-phase (I) and quadrature (Q) correlation values.

[0078] In this embodiment, these local copies are completely known and autonomously generated reference signals for the receiver processing unit. The specific implementation steps of the signal correlation processing are as follows:

[0079] First, the received signal is multiplied by two orthogonal local carriers to remove the carriers from the received signal, thus converting the signal back to baseband:

[0080] ;

[0081] ;

[0082] in, and These represent the baseband in-phase signal and quadrature signal after carrier stripping, respectively. It is the received baseband signal. and These are local in-phase carriers and quadrature carriers, respectively.

[0083] Next, the carrier stripped and The signals are respectively connected to three local codes ( E , P , L Multiply the signals and despread each signal to obtain the instantaneous correlation signal (e.g., ...). E Side roads, for P and L (The branch performs the exact same operation).

[0084] ;

[0085] ;

[0086] in, The (Prompt) code represents a local code copy that is phase-aligned with the currently best estimated received code. (Early) code and ( The codes represent the ratios respectively. A local copy of the code that is advanced and delayed by a certain number of code chips; and These represent the baseband in-phase and quadrature signals after early code despreading, respectively. It is the local Early code (E).

[0087] Finally, at a predetermined time interval Within this process, the results of the multiplication are summed to obtain the relevant result:

[0088] ;

[0089] ;

[0090] in, , This represents the in-phase and quadrature components of the early correlator. This represents the total number of sampling points within one integration period; It is a discrete-time index, representing the first... One sampling point; and They represent the first Discrete samples of the instantaneous correlation outputs of the early branch at each sampling time, where the outputs are in phase and orthogonal.

[0091] Similarly, performing the exact same operation on branches P and L yields the following result. (Integral result of branch P in phase) (Orthogonal integral results of branch P) and (Integral result of branch L in phase) (Orthogonal integral result of branch L).

[0092] S2. Multipath Detection: Analyze relevant results to identify and assess the presence and intensity of multipath interference.

[0093] In this embodiment, a multipath detection system is designed to detect the existence of multipath effects, and three complementary techniques are used to identify the presence and severity of multipath signals.

[0094] 2.1 Asymmetric analysis of morning and evening correlators:

[0095] The difference between the amplitudes of the complex output signals of the morning and evening correlators is used as a criterion for determining the presence of multipath interference. Under ideal signal conditions, the energy distribution of the morning and evening correlators is relatively symmetrical, meaning their amplitude ratio is stable. When multipath interference exists, the shape of the correlation peak shifts or stretches, resulting in asymmetry in the output amplitude of the morning and evening correlators.

[0096] The detection indicators are described as follows:

[0097] ;

[0098] in, , This represents the in-phase and quadrature components of the late correlator; the numerator of the detection index represents the amplitude difference between the early and late complex correlation signals, while the denominator normalizes the amplitude difference to eliminate amplitude deviations under different signal intensities. This represents the difference in amplitude between morning and evening, also known as the normalized phase detector output.

[0099] when When the value fluctuates within a very small range around 0, the correlation peak is symmetrical and there is no asymmetric distortion, indicating that there is no multipath or that there is negligible multipath; when When the correlation peak deviates significantly from 0, it is asymmetrical, indicating the presence of multipath interference.

[0100] 2.2 Code Error Variance Monitoring:

[0101] Ideally, the correlation peak of a delay-locked loop (DLL) is symmetrical. However, in a multipath environment, the direct path signal is superimposed with signals from delayed arrival paths such as reflections and diffractions, leading to correlation peak distortion. This distortion causes fluctuations in the code phase tracked by the DLL, manifesting as an increase in code error variance. When multipath is present, the DLL discriminator output is no longer the ideal value determined solely by the direct path, but rather expressed as:

[0102] ;

[0103] in, Indicates the total error. This indicates the error output of the phase detector for the direct signal itself in the absence of multipath propagation. Indicates the number of multipath signals. It is the amplitude ( It is the amplitude of the direct path signal. Indicates the first (amplitude of multipath signal) Represents the imaginary unit ( ), It is the code tracking error of the direct signal. Indicates the first The additional delay of a multipath signal relative to a direct signal. It is the first Phase shift of multipath signals ( (Carrier phase difference).

[0104] This formula outputs the discriminator of the main signal. The total error caused by multipath interference to the delay-locked loop (DLL) is monitored by superimposing the distorted outputs of all multipath signals (weighted by amplitude, phase, and time delay). .

[0105] In the above model, the output of the DLL depends not only on the main path response but also on the position, amplitude, and phase of the multipath components. Specifically, with the appearance and disappearance of multipath components, the DLL output error exhibits unstable random fluctuations. The variance of the code error can be expressed as:

[0106] ;

[0107] in, The variance of the total tracking error is represented by the following: This represents the variance of the error caused by noise. This represents the variance of the error caused by the multipath effect.

[0108] Through the By performing extensive statistical sampling and calculating the fluctuation (variance) of its output values, the above formula can be obtained. This formula decomposes the total error variance of the received signal into two independent components: noise and multipath interference, thus providing a theoretical benchmark for quantifying the severity of multipath interference. When multipath interference is present, The variance increases significantly, exceeding the variance level under pure noise conditions.

[0109] 2.3 Phase Locking Quality Assessment:

[0110] In multipath environments, phase-locked quality (measured as the ratio of in-phase power to total power) decreases, and its quantitative metric is as follows:

[0111] ;

[0112] in, It is a carrier tracking quality indicator that reflects the locking state of the carrier loop.

[0113] When the carrier phase is fully locked (phase error) At this time, almost all of the signal energy is concentrated in road, ,at this time This indicates the best locking quality; if the phase is not fully locked ( The signal energy will leak to The road led to Increase, at this time And the larger the phase error, The smaller.

[0114] A phase detector is used to measure the phase difference between an input signal and a local reference signal. In a multipath environment, its output can be expressed as:

[0115] ;

[0116] in, It is the phase error of the direct signal. It is the additional phase difference between the multipath signal and the direct signal. It is the carrier frequency of the GNSS signal.

[0117] This formula theoretically establishes a phase error model for multipath interference, but since the true phase of the direct signal cannot be directly obtained in a practical system... This formula is difficult to apply directly to multipath detection. Therefore, in practical processing, the received signal needs to be fed into a correlator to obtain the following branch output:

[0118] ;

[0119] ;

[0120] in, Indicates the direct path range. Indicates the amplitude of each multipath component. Indicates the number of multipaths.

[0121] The above expression shows that even when In the locked state, it is affected by multipath interference. It may still be significantly non-zero, thereby reducing the lock-in quality index ( ).

[0122] Furthermore, the multipath determination adopts a voting mechanism. First, it calculates... , , The detection results of these three indicators are used to determine the presence of multipath only when at least two indicators are identified as multipath (to improve detection reliability). Then, the severity of multipath is calculated. This serves as supplementary information to the detection output:

[0123] ;

[0124] in, , , These represent the quantified values ​​of multipath severity for early / late asymmetric analysis, code tracking error criterion, and phase-locked quality assessment, respectively. , , By performing simple mathematical processing on the results of these three formulas (taking the absolute value, calculating the variance, and calculating the deviation), we can obtain the final result. , , The weighting coefficients (0.4, 0.4, 0.2) are determined empirically based on the sensitivity and robustness of each indicator in a multipath environment, which can more confidently detect the results and severity of multipath signals.

[0125] S3. Basic Measurement Calculation: Using the I / Q output values ​​of the correlator, a specific phase detector algorithm is used to calculate the preliminary measurement error values, which represent the code phase lead / lag and carrier phase / frequency deviation, without complex filtering processing.

[0126] Based on the series of I / Q values ​​generated in step S1 ( , , , , , This allows for the calculation of preliminary code phase error and carrier phase error. The discriminator output is then used.

[0127] ;

[0128] in, For the instantaneous carrier phase error, it is fed into the carrier loop filter, adjusting the local carrier generator, ultimately making... When the value approaches zero, the loop locks. After locking, the accumulated change in the local carrier phase is the high-precision carrier phase measurement.

[0129] In this embodiment, an early / late power discriminator is used to calculate the code phase error based on the amplitude of the early and late branches. First, the power of the Early and Late branches is calculated:

[0130] ;

[0131] ;

[0132] in, and These represent the power of the Early and Late branches, respectively. Power calculations are performed to eliminate the influence of residual carrier phase, ensuring the discriminator is only sensitive to code phase.

[0133] Next, the discriminator output is calculated:

[0134] ;

[0135] in, For the instantaneous code phase error, it is fed into the code loop filter to adjust the local code generator, so that... When the value approaches zero, the code ring locks, and the local code phase after locking becomes the pseudorange measurement value.

[0136] S4. Measurement Enhancement: For the original carrier phase measurement values ​​obtained from the basic measurement calculation in step S3, a carrier phase smoothing algorithm based on historical Doppler data prediction and anomaly detection is applied to reduce its short-term drastic fluctuations, providing a more stable and reliable measurement input for subsequent adaptive Kalman filtering and improving the quality of the innovation sequence.

[0137] S5. Dynamic adjustment of noise covariance: Based on multipath detection, in order to reduce the Kalman filter's trust in unreliable data, a dynamic adjustment mechanism for noise covariance is introduced.

[0138] This method aligns with the Bayesian estimation principle of increasing the noise covariance to reduce the impact of observations when information uncertainty increases. Noise covariance matrix adjustment involves revising both the measurement noise covariance and process noise covariance matrices.

[0139] First, define code phase noise. Carrier phase noise The initial measurement noise covariance matrix is ​​obtained as follows:

[0140] ;

[0141] The first row and column correspond to code phase noise, and the second row and column correspond to carrier phase noise. Theoretically, multipath interference usually has a more significant impact on code tracking than on carrier tracking because the code phase has a lower rate of change and is more susceptible to interference from multipath signals.

[0142] Then, based on the multipath severity detected in step S2 ( Define the code phase noise gain coefficient and the carrier phase noise gain coefficient respectively:

[0143] ;

[0144] ;

[0145] in, and These represent the code observation noise scaling factor and the carrier phase observation noise scaling factor, respectively.

[0146] Next, for the matrix Dynamically scale the main diagonal elements:

[0147] ;

[0148] ;

[0149] in, It is the first row and column element of the initial measurement noise covariance matrix. That is, code phase noise. It is the second row and column element of the initial measurement noise covariance matrix. This refers to carrier phase noise. and Representation matrix The scaled value of the main diagonal element.

[0150] Finally, the adjusted covariance matrix Divide by cycle time To obtain the final measurement noise covariance :

[0151] ;

[0152] Furthermore, for the process noise covariance matrix Adjustments will also be made to address different multipath effects. The gain coefficient for each state is based on... Make dynamic adjustments:

[0153] Regarding the code phase state:

[0154] ;

[0155] Regarding carrier phase state:

[0156] ;

[0157] Regarding frequency states:

[0158] ;

[0159] For the rate of change of frequency:

[0160] ;

[0161] in, , , , The process noise covariance matrix The four main diagonal elements, , , , For matrix The scaled values ​​of the four main diagonal elements.

[0162] In the scaled matrix Based on this, the discretization form of the state transition matrix is ​​used. and time-related change matrix The discretized process noise covariance matrix at the current time is obtained. :

[0163] ;

[0164] The superscript T indicates transpose.

[0165] S6. Adaptive Kalman Filter: Combining the dynamically adjusted noise model, the state prediction and update recursive calculation of the Kalman filter are performed. At the same time, adaptive algorithms such as Sage-Husa are used to estimate and correct noise statistical parameters online based on the filtering information, thereby obtaining the optimal estimate of key states such as signal carrier phase, frequency, and code phase.

[0166] S6.1. Using Kalman filtering combined with the Sage-Husa algorithm, noise parameters are dynamically adjusted to estimate the carrier phase, frequency, and code phase state of the signal.

[0167] In the Kalman filtering process, the process noise covariance and observation noise covariance It will be updated in real time, so it will , Replace with , Specifically, the Sage-Husa adaptive filtering algorithm is used to apply the current measurement residual, i.e., the innovation, to... and Adjustments will be made.

[0168] New sequence The expression is:

[0169] ;

[0170] in, yes The actual observation vector obtained by the receiver at each moment is a value that has been smoothed in step S4, resulting in less random noise and a smaller, smoother residual. It is the predicted observation vector obtained by mapping the predicted state from the previous time step through the observation model. It is the optimal state estimate from the previous moment. It is the observation matrix.

[0171] Let the new information sequence be The variance is :

[0172] ;

[0173] in, This represents the expectation operator.

[0174] In a discrete-time stochastic linear system, system noise and observation noise are denoted as process noise, respectively. and observation noise Its statistical properties exhibit time-varying characteristics, specifically the time-varying nature of its mean and covariance. The system satisfies the following statistical property constraints:

[0175] ;

[0176] in, and These are time indices; these two variables are used to describe the correlation between noisy processes at different points in time. It is the process noise of the system. The mean (mathematical expectation) of. It is observation noise The mean (mathematical expectation) of. Describes the covariance operator. Indicates another moment Process noise, Indicates another moment Observation noise.

[0177] It is the Kronecker delta function, defined as:

[0178] ;

[0179] Therefore, and All of them are independent zero-mean white noise sequences.

[0180] Specifically, this embodiment employs a dynamic update mechanism based on the Sage-Husa adaptive filter to update the current time... process noise mean Covariance Observational noise mean Covariance Perform recursive estimation. Combine innovation with Kalman gain. Update the noise statistics:

[0181] ;

[0182] ;

[0183] ;

[0184] ;

[0185] in, A fixed decay factor is used to balance the impact of new and old data on the current estimate. The specific expression is:

[0186] ;

[0187] In the formula, , It is a forgetting factor.

[0188] The estimated noise statistics ( , , , The noise covariance is fed back to the main loop of the Kalman filter in real time, and is adaptively corrected by adjusting the process noise covariance. ) and observation noise covariance ( The parameters are dynamically optimized to improve the convergence characteristics and robustness of the filter, so as to cope with the uncertainty of dynamic models and observation environment, and ultimately improve the overall estimation accuracy in complex scenarios.

[0189] S6.2 replaces the fixed attenuation factor in the adaptive Kalman filter with an adaptive factor that is dynamically adjusted according to the severity of multipath interference, thus accurately tracking the carrier phase.

[0190] To further enhance the system's adaptability to multipath interference, this embodiment modifies the traditional fixed attenuation factor. Improvements were made by introducing an adaptive factor that can dynamically adjust based on the severity of multipath attacks. to replace Its function is to achieve adaptive control of the Kalman filter update rate. Its theoretical basis is:

[0191] ;

[0192] in, It is an adaptive factor that is dynamically adjusted based on the severity of multipath attacks; These are the base values ​​used to ensure initial convergence; This is an incremental function related to the severity of multipath propagation, used to dynamically adjust the update speed based on relevant metrics provided by the multipath detection module. It does not have a single standard form, but its expression is explicit and can be designed according to requirements (e.g., two-segment, exponential, three-segment functions, etc.).

[0193] In highly dynamic navigation scenarios (such as GNSS signal tracking), the initial state estimation of the system has significant uncertainties. Introducing adaptive updates of the noise covariance prematurely before the filter converges may lead to significant deviations or divergence in the state estimation. Therefore, the following adaptive adjustment mechanism is adopted to ensure good convergence characteristics in the initial state while also considering the responsiveness to multipath interference.

[0194] First, within the first 50ms of system initialization (corresponding to the number of tracking loops) Using a fixed default noise covariance matrix without adaptive adjustment ensures that the filter converges quickly to the true state of the system without interference.

[0195] Secondly, when the multipath detection subsystem determines that multipath interference exists in the area, the system adjusts the response based on the severity of the multipath interference and the tracking time. Adjustments were made accordingly. By introducing a basic update rate parameter, it was ensured that even under weak disturbances or mild multipath environments, the system noise was still dynamically estimated to a minimum, effectively improving the system's adaptability and stability in edge states and avoiding the problem of completely static covariance updates.

[0196] The specific calculation process is as follows:

[0197] ;

[0198] ;

[0199] in, The base update rate is typically set to [0.1, 0.3], with a preferred value of 0.2 in this embodiment to balance the filter convergence speed with the noise estimation sensitivity.

[0200] This is a time decay factor used to control the decay rate of multipath adjustment, thereby improving system stability and avoiding long-term over-response.

[0201] It is the number of filtering iterations. As an additional adjustment rate, in the initial stage ( The more severe the multipath effect, the better. The larger, The larger; in the subsequent stages ( (times), to Adjustments are made to arrive at the final update rate. .

[0202] Final adaptive factor The calculation is as follows:

[0203] ;

[0204] Here, 0.95 is an adaptive factor to avoid noise covariance estimation. The maximum allowable weight factor is set to prevent excessively large weights from causing drastic fluctuations in the filter update weights.

[0205] When the system is in an environment without significant multipath interference, the adaptive factor is adopted. The function is based on exponential decay over time, and its specific expression is as follows:

[0206] ;

[0207] The attenuation factor in adaptive Kalman filtering is fixed or slowly varying. Replace with an adaptive factor that is dynamically adjusted based on the severity of multipath attacks. This enables precise tracking of the carrier phase, and this dual adaptive mechanism can improve filtering accuracy and robust stability.

[0208] S7. Tracking Parameter Update: The control parameters of the receiver's internal tracking loop are updated using the estimation results of the Kalman filter.

[0209] In this embodiment, the noise statistics estimated using a Kalman filter ( , This allows for dynamic adjustment of loop filter parameters (such as bandwidth).

[0210] Increasing the loop bandwidth indicates a poor observation environment (severe multipath propagation, low signal-to-noise ratio), requiring a reduction in loop bandwidth. To suppress noise; Increasing the loop bandwidth indicates that the receiver has high dynamics (such as acceleration and turning), requiring a larger loop bandwidth. To quickly track signal changes.

[0211] Specifically, adaptive algorithms (such as lookup tables, fuzzy functions, or optimization functions) are designed based on... and Real-time calculation of the optimal noise equivalent bandwidth of the carrier loop and code loop ( ):

[0212] ;

[0213] Among them, the function It is used to dynamically calculate the optimal filter bandwidth or control parameters based on real-time estimates of system noise and measurement noise.

[0214] Furthermore, the calculated This is converted into the coefficients of the carrier loop and code loop filters. Changes in the filter coefficients directly determine the loop-to-discriminator error. , The response speed and smoothness of the signal are improved, thereby indirectly optimizing the adjustment process of the carrier frequency and code frequency. As phase accumulation quantities, the accuracy and stability of the accumulated carrier phase and code phase are also significantly improved.

[0215] S8. Data storage and iteration: Save the tracking results and internal state information of the filter calculated in the current tracking cycle to ensure that the processing flow can be seamlessly connected and run continuously in the next cycle based on the current results, thereby realizing carrier loop multipath suppression of Kalman filtering.

[0216] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A carrier loop multipath suppression method based on an improved Sage-Husa adaptive Kalman filter, characterized in that, Includes the following steps: Step 1, Signal Correlation Processing: Generate three sets of local reference signals with early, positive, and late time offsets, perform correlation operations with the input satellite signals respectively, and extract the time delay error using the in-phase and quadrature outputs to obtain the correlation results of the early, positive, and late channels; Step 2, Multipath Detection: Analyze the ratio of the correlation results of the early and late paths to determine the phase lock quality and code phase tracking fluctuations. Use a multipath detection system to detect whether multipath effects exist, identify and assess the existence and severity of multipath interference. Step 3, Basic Measurement Calculation: Calculate the code phase error using the correlation outputs of the early and late paths, and calculate the preliminary carrier phase tracking error using the phase relationship between the in-phase and quadrature components of the positive branch; Step 4, Measurement Enhancement: The original carrier phase measurement value is smoothed and automatically adjusted according to the severity of multipath, reducing the phase error caused by noise and multipath effects before entering the filter; Step 5, Dynamic adjustment of noise covariance: Based on the severity of the detected multipath, dynamically adjust the Kalman filter's prediction of the measurement noise covariance and process noise covariance; Step 6: Adaptive Kalman Filtering: Using Kalman filtering combined with the Sage-Husa algorithm, noise parameters are dynamically adjusted to estimate the carrier phase, frequency, and code phase state of the signal; the fixed attenuation factor in the adaptive Kalman filter is replaced with an adaptive factor that is dynamically adjusted according to the severity of multipath, so as to accurately track the carrier phase. Step 7, Tracking Parameter Update: Update the control parameters of the receiver's internal tracking loop using the estimation results of the Kalman filter; Step 8, Data Storage and Iteration: Save the processing results and status of the current cycle to prepare for the next round of tracking calculation, and realize carrier loop multipath suppression of Kalman filtering.

2. The carrier loop multipath suppression method according to claim 1, characterized in that, In step 1, the signal correlation processing includes: Step 1.1: Receive satellite signals With two orthogonal local carriers and Multiply the signal, remove the carrier from the received signal, and convert the signal back to baseband to obtain the baseband in-phase signal after carrier stripping. signals orthogonal to baseband ; Step 1.2: The baseband in-phase signal after carrier stripping signals orthogonal to baseband The signals are multiplied by the three local reference signals of morning, mid-day, and evening respectively, and each signal is despread to obtain the three baseband in-phase and quadrature signals of morning, mid-day, and evening respectively. Step 1.3, at the predetermined time interval Within this process, the early, positive, and late baseband in-phase and quadrature signals after despreading are accumulated separately to obtain the three-channel correlation results, which are expressed as follows: , , , , , ; Among them, subscript , , These correspond to the three local reference signal code copies for morning, noon, and evening respectively. , for Results of in-phase and orthogonal integrals of branches , for Results of in-phase and orthogonal integrals of branches , for Results of in-phase and orthogonal integrals of the branches.

3. The carrier loop multipath suppression method according to claim 2, characterized in that, In step 2, the multipath detection system employs three complementary methods to identify the presence and severity of multipath signals, including: Step 2.1: Morning and Evening Asymmetric Analysis. The difference in the amplitude of the complex signal output from the morning and evening correlation results is used as the basis for determining the presence or absence of multipath interference. ； in, The numerator represents the amplitude difference between the two complex correlation signals in the morning and evening, while the denominator normalizes the amplitude difference to eliminate amplitude deviation under different signal intensities. Step 2.2, Code error variance monitoring, the variance of the code error is expressed as: ； in, The variance of the total tracking error is represented by the following: This represents the variance of the error caused by noise. This represents the error variance caused by multipath effects; when multipath exists, The variance increases, exceeding the variance level in a purely noisy environment; Step 2.3, Phase-locking quality assessment, with the following quantitative indicators: ； in, It is a carrier tracking quality indicator used to reflect the locking state of the carrier loop: when the carrier phase is fully locked, the signal energy is concentrated in the carrier loop. road, ,at this time This indicates optimal lock quality; if the phase is not fully locked, signal energy leaks out. The road led to Increase, at this time And the larger the phase error, The smaller.

4. The carrier loop multipath suppression method according to claim 3, characterized in that, The multipath effect is determined using a voting mechanism, which calculates the detection results of three indicators from three complementary methods. Multipath is considered to exist when at least two indicators indicate multipath propagation, and the severity of multipath propagation is then calculated. This serves as supplementary information to the detection output: ； in, , , These represent the multipath severity quantification values ​​for early and late asymmetric analysis and code error variance monitoring, i.e., phase-locked quality assessment.

5. The carrier loop multipath suppression method according to claim 3, characterized in that, In step 3, the code phase error is initially calculated using the correlation outputs from the morning and evening paths. : ； in, and These represent the power of the morning and evening branches, respectively: ； ； The preliminary carrier phase tracking error is calculated using the phase relationship between the in-phase and quadrature components of the positive branch. : 。 6. The carrier loop multipath suppression method according to claim 5, characterized in that, In step 5, the measurement noise covariance matrix is ​​adjusted as follows: Step 5.11: Define the initial measurement noise covariance matrix. : ； in, Indicates code phase noise, Indicates carrier phase noise; Step 5.12: Based on the multipath severity detected in step S2, define the code phase noise gain coefficient and the carrier phase noise gain coefficient respectively, and apply them to the matrix. Dynamically scale the main diagonal elements: ； ； in, and These represent the code observation noise scaling factor and the carrier phase observation noise scaling factor, respectively. and Representation matrix The scaled value of the main diagonal element; Step 5.13: Divide the scaled covariance matrix by the loop time to obtain the final measurement noise covariance. : ； in, This is the scaled covariance matrix. The loop time is [time].

7. The carrier loop multipath suppression method according to claim 6, characterized in that, In step 5, the process noise covariance matrix is ​​adjusted as follows: Step 5.21: Define the initial process noise covariance matrix. ; Step 5.22: Based on the multipath severity detected in step S2, define the code phase state gain coefficient, carrier phase state gain coefficient, frequency state gain coefficient, and frequency change rate state gain coefficient, respectively. Then, apply the method from step 5.12 to the matrix. The diagonal elements are dynamically scaled to obtain a scaled matrix. ; Step 5.23, based on the scaled matrix Through the discretized form of the state transition matrix and time-related change matrix The discretized process noise covariance matrix at the current time is obtained. : ； The superscript T indicates transpose.

8. The carrier loop multipath suppression method according to claim 7, characterized in that, In step 6, Kalman filtering combined with the Sage-Husa algorithm is used to dynamically adjust the noise parameters, combining innovation and Kalman gain. Regarding the current moment process noise mean Covariance Observational noise mean Covariance Perform recursive estimation; ； ； ； ； in, Indicates the previous moment The estimated value of the mean process noise, , Indicates the current time The previous moment State estimates, Here is the state transition matrix. A fixed decay factor is used to balance the impact of new and old data on the current estimate; It was the previous moment The estimated value of the process noise covariance matrix, It is the current moment. The posterior estimation error covariance matrix, It was the previous moment The posterior estimation error covariance matrix, Indicates the receiver at the current time. The observation vector formed during loop update is used to characterize the code alignment error observation and carrier phase alignment error observation of the current loop; It is the observation matrix; Indicates the previous moment The estimated value of the mean measurement noise. This represents the prior estimate of the state vector. Indicates the previous moment The estimated value of the measurement noise covariance, It is the prior estimation error covariance.

9. The carrier loop multipath suppression method according to claim 8, characterized in that, In step 6, the fixed attenuation factor in the adaptive Kalman filter is... Replace with an adaptive factor that is dynamically adjusted based on the severity of multipath attacks. When multipath interference is detected, adjustments are made based on the severity of the multipath interference and the tracking time. Adjustments are made to the adaptive factor. The calculation is as follows: ； Where 0.95 is the maximum allowable weighting factor; The calculation process for the update rate is as follows: ； ； in, The time decay factor, This represents the number of filtering iterations. Additional adjustment rate; Based on the update rate, Multipath severity; In the initial stage, the more severe the multipath effect, The larger, The larger; in subsequent stages, for Adjustments are made to arrive at the final update rate. .

10. The carrier loop multipath suppression method according to claim 8, characterized in that, In step 7, an adaptive algorithm is used to determine the noise statistics estimated using the Kalman filter. and Real-time calculation of the optimal noise equivalent bandwidth of the carrier loop and code loop. : ； Among them, the function Used to dynamically calculate the optimal filter bandwidth or control parameters based on real-time estimates of system noise and measurement noise; the calculated... The coefficients are converted into carrier loop and code loop filter coefficients, which determine the loop's response speed and smoothness to discriminator errors, thus optimizing the adjustment process of the carrier frequency and code frequency.

Images