Satellite denial environment positioning method based on multi-source opportunity signal fusion

Abstract

The application discloses a satellite denial environment positioning method based on multi-source opportunity signal fusion, and relates to the technical field of satellite navigation and multi-source fusion positioning. In view of the sparse bias caused by the multipath effect in the urban canyon and the variance and scale mismatching problem existing in the GNSS, 5G and DTMB heterogeneous signal fusion, the application firstly constructs a unified heterogeneous observation model and explicitly introduces a sparse bias term; secondly, a diagonal pre-whitening matrix is constructed by using a measurement noise covariance matrix to multiply a residual error equation, so that the heterogeneous signals are unified to a comparable statistical scale; then, a projection matrix is introduced in the pre-whitening domain to eliminate state increment interference, and L1 regularization is combined to accurately solve a sparse bias estimation value; finally, the bias is deducted from the original observation to obtain pure innovation, and an extended Kalman filter is input to complete state updating. Without directly eliminating abnormal links, the application adaptively suppresses the non-line-of-sight bias, and significantly improves the positioning precision and robustness in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of satellite navigation and multi-source wireless signal fusion positioning technology, specifically to a fusion positioning method for urban canyons and satellite navigation-constrained environments, and particularly to a method for robust positioning by constructing a unified pseudorange and pseudorange rate observation model using heterogeneous measurements of GNSS, 5G and DTMB, and combining diagonal pre-whitening, sparse bias estimation and extended Kalman filtering. Background Technology

[0002] Global Navigation Satellite Systems (GNSS) offer advantages such as wide coverage and low deployment costs. However, in urban canyon environments, they are susceptible to building obstruction, reflection, and non-line-of-sight (NLS) propagation, introducing significant multipath and NLS errors into pseudorange and Doppler observations. These errors typically do not affect all links simultaneously but rather manifest as sparse deviations occurring on a small number of links and at a limited number of moments, leading to drift or even instability in traditional filtered positioning results.

[0003] When GNSS availability declines, opportunistic signals such as 5G and DTMB can provide additional geometric constraints and observation redundancy. Therefore, the joint fusion of GNSS with 5G and DTMB has become an important direction for improving positioning availability and continuity in complex environments. However, most existing urban environmental error suppression methods adopt strategies such as empirical weighting, fixed threshold elimination, multiple hypothesis testing, or integrity monitoring. These methods usually rely on strong observation redundancy. In the weak observability environment of urban canyons, directly eliminating some abnormal observations often further weakens the geometric configuration, resulting in decreased observability, increased false positives and false negatives, and increased risk of filter divergence.

[0004] Furthermore, GNSS, 5G, and DTMB are heterogeneous measurement sources, and different systems differ in noise levels and other aspects. If a uniform sparse estimation model is directly applied to all residuals in multi-source fusion without considering heteroscedasticity, the objective function may become unbalanced, with high-variance or large-scale residuals dominating the estimation process. This can lead to inconsistent sparse decision thresholds, excessive bias contraction, or missed detections, and reduce the consistency and numerical stability of the fusion filter.

[0005] To address the aforementioned problems in existing technologies, this invention provides a GNSS / 5G / DTMB fusion positioning method based on diagonal pre-whitening sparse bias estimation. This method is used in urban canyons and GNSS-constrained environments to adaptively estimate and suppress link-level anomalies caused by multipath and non-line-of-sight biases, thereby improving positioning accuracy, stability, and availability while preserving effective observations. Summary of the Invention

[0006] The technical problem to be solved by this invention is achieved through the following technical solution:

[0007] (1) Construct a unified GNSS / 5G / DTMB observation model: unify the pseudorange and pseudorange rate measurements of each system, and explicitly introduce a sparse bias term into the observation residual equation to characterize the abnormal error.

[0008] (2) Perform diagonal pre-whitening transformation: Use the measurement noise covariance matrix to pre-whiten the residual equation, unifying the heterogeneous signals to a comparable statistical scale.

[0009] (3) Solve and estimate sparsity bias: In the pre-whitening domain, the sparsity bias caused by multipath or NLOS is calculated using the projection matrix and an optimization algorithm based on L1 regularization.

[0010] (4) Perform bias correction and filter update: remove the estimated sparse bias from the original observation residuals and update the state of the extended Kalman filter (EKF) using the corrected pure information.

[0011] Furthermore, the unified observation model described in step (1) of this invention is a vector composed of all pseudoranges and pseudorange rate residuals, and its residual equation is expressed as:

[0012] in For geometric observation matrix, For state increment prediction error, It is a sparse bias vector. To measure the noise vector.

[0013] Furthermore, the diagonal pre-whitening transformation in step (2) of this invention refers to constructing a block diagonal covariance matrix composed of the measurement variances of each link. And multiply the pre-whitening matrix on the left side of both sides of the residual equation: This prevents high variance or large-scale measurements from dominating the estimation process.

[0014] Furthermore, the calculation of the sparsity bias in step (3) of this invention is achieved by constructing an objective function of the following form for solving:

[0015] Using projection matrix Filter out components that are indistinguishable from system status updates.

[0016] Furthermore, the multi-source fusion localization algorithm in step (4) of this invention refers to calculating the pure information:

[0017] The pure information is then substituted into the Kalman gain formula to complete the final unbiased estimation of the receiver position, velocity, and clock bias.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention effectively solves the scale mismatch and heteroscedasticity problems in heterogeneous fusion of GNSS, 5G, and DTMB through diagonal pre-whitening, thus improving numerical conditions. Simultaneously, without directly eliminating anomalous observation links, this invention adaptively estimates and suppresses sparsely distributed multipath and non-line-of-sight biases, significantly improving positioning accuracy, continuity, and algorithm stability in complex urban environments. Attached Figure Description

[0019] Figure 1 This is a flowchart of a satellite denial environment positioning method based on multi-source opportunistic signal fusion according to the present invention.

[0020] Figure 2 This is a comparison chart of the mean square error (MSE) of positioning when abnormal pseudorange deviation is introduced in this invention. Detailed Implementation

[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. The drawings are provided only to assist in understanding the present invention and do not constitute a limitation thereof. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0022] See appendix Figure 1 The implementation steps for this example are as follows: (1) Construct a unified heterogeneous observation model; The system acquires pseudorange measurements and pseudorange rate measurements for all available links in the GNSS, 5G, and DTMB systems at the current epoch. It defines the system's state vector at the current epoch, which is composed of the receiver's three-dimensional position components, receiver clock offset, three-dimensional velocity components, and clock drift.

[0023] For each observation link, pseudorange observation equations and pseudorange rate observation equations are established separately. A sparse bias term is added to these equations. The prediction residuals (i.e., innovation) for each link are calculated, and all pseudorange and pseudorange rate residuals are stacked in a fixed order to generate a unified residual vector. The stacked residual equation is expressed as follows:

[0024] in, The residual vectors are stacked. This is a block diagonal geometric observation matrix composed of unit line-of-sight vectors for each link. For state increment prediction error, This is the stacked sparse bias vector. To measure the noise vector.

[0025] (2) Diagonal pre-whitening transformation; The system extracts the prior measurement noise variance for each GNSS, 5G, and DTMB observation link at the current epoch. Using these variance values ​​as diagonal elements, a block diagonal measurement noise covariance matrix with a size matching the stacked residual vector is constructed, denoted as... .

[0026] Calculate the negative 1 / 2 power of the covariance matrix to obtain the diagonal pre-whitening matrix. Multiply this pre-whitening matrix on the left of both sides of the residual equation constructed in step 1. Based on the pre-whitened residual equation, and combined with the L1 regularization term, construct a joint optimization objective function that includes state increments and sparse biases:

[0027] (3) Decoupling and estimation of sparse bias; The system calculates and generates a projection matrix based on the geometric observation matrix and the pre-whitening matrix. The projection matrix is ​​then applied to the objective function constructed in step 2 to eliminate the state increment prediction error variable, transforming the original objective function into one containing only a sparse bias vector. The multi-source fusion positioning algorithm is designed in the form of [the algorithm], and the performance of the positioning system is evaluated. By solving this L1 regularization optimization problem, the system outputs the sparse bias estimation vector for the current epoch.

[0028] (4) Bias correction and extended Kalman filter update; The system will obtain the sparse bias estimation vector from step 3. Substituting it into the corrected equation, we can extract it from the original stacked residual vector. Subtract from the middle. Output the clean innovation vector after removing sparse bias. .

[0029] Subsequently, the system enters the extended Kalman filter update phase. The system calls the predicted state covariance matrix, geometric observation matrix, and measurement noise covariance matrix, and calculates the innovation covariance matrix and Kalman gain in sequence.

[0030] Finally, the system uses the Kalman gain and the pure innovation vector, superimposed with the predicted state vector, to calculate the final state estimate for the current epoch. :

[0031] like Figure 2 The figure shown is a comparison of the mean square error (MSE) of the positioning method in the embodiment of the present invention and the traditional standard extended Kalman filter (EKF) method when encountering severe multipath interference.

[0032] In experimental scenarios, when significant pseudorange biases (e.g., 60 meters) are injected into parts of the GNSS link, the traditional EKF method, unable to isolate anomalous observations, directly absorbs systematic errors, leading to a rapid amplification of positioning errors. In contrast, the fusion positioning method described in this invention unifies the heteroscedasticity of multi-source signals through diagonal pre-whitening and accurately removes sparsity biases. Significantly, the SparseEKF curve remains stable near the zero mark throughout the entire multipath interference range, exhibiting minimal fluctuations. This demonstrates that the proposed method of explicitly estimating and removing structured biases effectively suppresses systematic errors caused by multipath effects, significantly improving the robustness of the fusion positioning system. After the epoch interference ends, although both methods can recover, SparseEKF demonstrates significantly greater stability across the entire trajectory.

[0033] The above embodiments are only used to illustrate the present invention. All equivalent transformations and improvements made based on the technical solutions of the present invention should be considered within the protection scope of the present invention.

Claims

1. A satellite-denied environment positioning method based on multi-source opportunistic signal fusion, wherein the method is executed by a navigation and positioning receiver or related processing equipment, characterized in that, Includes the following steps: Step 1: Obtain the measurement values ​​of each available heterogeneous link in the current epoch, and construct a unified residual equation model that includes state increment prediction error, sparse bias vector and measurement noise vector; Step 2: Extract the prior measurement noise variance corresponding to each observation link to construct the measurement noise covariance matrix, perform diagonal pre-whitening transformation on the unified residual equation model, and construct an objective function with state increment prediction error and sparsity bias as the variables to be solved. Step 3: Based on the objective function, the projection matrix is ​​used to eliminate the state increment prediction error variable. By solving the L1 regularization optimization problem, the sparse bias vector is decoupled and estimated. Step 4: Subtract the estimated sparse bias vector from the original residual to obtain the corrected pure innovation vector. Use the pure innovation vector to perform extended Kalman filter state update and output the localization solution for the current epoch.

2. The method according to claim 1, characterized in that, The specific process of step 1 is as follows: Obtain pseudorange measurement values ​​and pseudorange rate measurement values ​​from GNSS, 5G, and DTMB systems, and establish pseudorange observation equations and pseudorange rate observation equations containing sparse bias terms respectively; calculate the prediction residuals of each link, and stack all pseudorange residuals and pseudorange rate residuals to generate a unified residual vector; this residual vector is represented as the product of the geometric observation matrix composed of the line-of-sight unit vectors of each link and the state increment prediction error, plus the sum of the stacked sparse bias vector and the measurement noise vector.

3. The method according to claim 2, characterized in that, The specific process of step 2 is as follows: Using the extracted prior measurement noise variance values ​​of each link as diagonal elements, construct the measurement noise covariance matrix of the block diagonal; calculate the negative 1 / 2 power of the measurement noise covariance matrix, use it as the diagonal pre-whitening matrix and multiply it on the left to both sides of the unified residual equation; combine the L1 regularization term to construct a single objective function with the state increment prediction error and sparse bias as the variables to be solved. The objective function includes the L2 norm squared term of the pre-whitening residual and the L1 norm penalty term of the sparse bias vector.

4. The method according to claim 3, characterized in that, The specific process of step 3 is as follows: Calculate and generate an orthogonal projection matrix using the geometric observation matrix and the pre-whitening matrix; apply the orthogonal projection matrix to the single objective function, filter out components related to normal state updates to eliminate state increment prediction error variables, and convert the single objective function into an objective function containing only sparse bias vectors; solve the optimization problem of minimizing the objective function to calculate the sparse bias estimation vector of the current epoch.

5. The method according to claim 4, characterized in that, The specific process of step 4 is as follows: Subtract the sparse bias estimation vector from the original unified residual vector to obtain a pure innovation vector that has eliminated abnormal errors; call the extended Kalman filter and calculate the innovation covariance matrix and Kalman gain in sequence using the predicted state covariance matrix, geometric observation matrix and measurement noise covariance matrix; use the Kalman gain to weight the pure innovation vector and superimpose it on the predicted state vector of the system to obtain the final state estimate of the current epoch, and at the same time complete the update of the state covariance matrix.

Images