CN-121720491-B - Underwater navigation self-adaptive filtering method based on Keramelteon lower bound constraint

CN121720491BCN 121720491 BCN121720491 BCN 121720491BCN-121720491-B

Abstract

The invention discloses an underwater navigation self-adaptive filtering method based on a Keramelteon lower bound constraint, which belongs to the technical field of underwater navigation positioning and is used for underwater navigation and comprises the steps of modeling non-Gaussian measurement noise as a Gaussian scale mixed model aiming at a discrete time system; obtaining a robust innovation sequence and a weighting factor through variation iterative updating, which are used for correcting the measurement information matrix and updating the posterior Kramer lower bound; the empirical estimation and the Kelarmew lower bound-adaptive adjustment process noise covariance matrix are combined to optimize the state prediction. According to the invention, a closed-loop feedback and decision mechanism with the lower Cramerro bound as a theoretical performance scale is introduced, so that the self-adaptive process is estimated from an open loop depending on empirical data, and intelligent optimization of closed-loop anchoring and calibration of an acceptance theory optimality criterion is broken through, thereby realizing precise decoupling and coordination of time-varying noise tracking and abnormal interference suppression, and finally achieving performance improvement of underwater high-precision steady navigation positioning.

Inventors

WANG LEI
SUN TONG
ZHANG JIA
WANG SHENGLI
Niu Xiaoman
LIU YUHENG

Assignees

山东科技大学

Dates

Publication Date: 20260508
Application Date: 20260227

Claims (10)

1. The underwater navigation self-adaptive filtering method based on the lower bound constraint of the Keramelteon is characterized by comprising the following steps of: S1, establishing a state equation and an observation equation of a discrete time system, expressing measurement noise as a Gaussian scale mixed model, and expressing an edge probability density function of the measurement noise as an integral form of conditional Gaussian distribution and gamma distribution based on the Gaussian scale mixed model; S2, predicting a state vector and a state covariance matrix by using a process noise covariance matrix, setting an iteration frequency threshold at each moment, and carrying out variation iteration update, wherein the variation iteration update sequentially comprises updating an innovation covariance matrix, calculating the expectation of the square of a measurement residual error and updating a weighting factor of an auxiliary variable, and obtaining a final robust innovation sequence and a final weighting factor after reaching the iteration frequency threshold; s3, predicting an information matrix by using the Fisher information matrix, calculating a non-Gaussian corrected measurement information matrix by using a final weighting factor, and updating a posterior Kramer lower bound by using the predicted information matrix and the non-Gaussian corrected measurement information matrix; S4, calculating an empirical estimation value of a process noise covariance matrix at the previous moment by utilizing a final robust information sequence, then combining improved Sagnac Hu Sazi to adapt filtering, calculating the empirical estimation value of the process noise covariance matrix at the current moment, constructing a constraint tolerance based on a Keramer lower bound, calculating a process noise covariance matrix subjected to tolerance constraint processing based on the constraint tolerance and the empirical estimation value of the process noise covariance matrix at the current moment, returning the process noise covariance matrix subjected to tolerance constraint processing to the step S2, and predicting a state vector at the next moment.
2. The adaptive filtering method for underwater navigation based on the lower bound constraint of claimepir according to claim 1, wherein S1 comprises, S1.1, obtaining the triaxial acceleration and the angular velocity of an inertial navigation system, and calculating the motion state of an autonomous underwater vehicle; measuring the distance between the autonomous underwater vehicle and the acoustic beacon by using a long baseline positioning system through arranging the acoustic beacon at a known position on the seabed, and determining the position of the autonomous underwater vehicle through geometric intersection; Sequentially selecting speed error, attitude error, position error, accelerometer zero offset and gyro drift, and defining state vector in northeast navigation coordinate system ; ; Based on an inertial navigation system and a long baseline positioning system, a state equation and an observation equation of a discrete time system are established: ; in the formula, For the residuals of the long baseline measurement and inertial navigation system prediction, In the form of a state transition matrix, In order to measure the matrix of the device, For the noise to drive the matrix, In order for the process to be noisy, Mean value 0 covariance Is provided for the distribution of (a), Is that Process noise covariance matrix of time of day, In order to measure the noise of a person, Is the time of day.
3. The adaptive filtering method for underwater navigation based on the lower bound constraint of Keramelteon as claimed in claim 2, wherein S1 comprises S1.2, constructing a hierarchical probability model of non-Gaussian noise, and setting Obeys Student-t distribution and will Expressed as a gaussian scale mixture model, a hierarchical structure is defined: ; ; in the formula, In order to assist in the random variable, Obeys the gamma distribution, In order for the degree of freedom to be provided, In order to follow the distribution symbol, In order to measure the noise covariance matrix, In the form of a gaussian distribution, Is gamma distribution; based on the layered structure, will Is the edge probability density function of (2) Expressed as an integral of the conditional gaussian distribution and the gamma distribution: 。
4. The adaptive filtering method for underwater navigation based on the lower bound constraint of claim 3, wherein S2 comprises S2.1, using Process noise covariance of time of day Standard time updates are performed: ; ; in the formula, Is that The predicted state vector of the moment in time, Is that The posterior state estimate vector of time of day, Is that The prediction state covariance matrix of the moment in time, Is that Time posterior state covariance matrix.
5. The adaptive filtering method for underwater navigation based on the lower bound constraint of claim 4, wherein S2 comprises S2.2, wherein S2 is the following weight percentage Setting iteration times threshold at moment Performing A second fixed-point iteration including updating the innovation covariance : ; ; In the formula, Is that The equivalent measured noise covariance matrix in the ith iteration of the time variant iteration, For the number of iterations, , Is that Time of day (time) The weighting factors of the auxiliary variables of the second iteration, Transpose the symbol; calculating the expectation of the square of the measurement residual : ; In the formula, Trace operator for matrix; Updating the weighting factors of the auxiliary variables: ; in the formula, In order to measure the dimension of the dimension, Is that Time of day (time) Weighting factors for auxiliary variables of the multiple iterations.
6. The adaptive filtering method for underwater navigation based on the lower bound constraint of Keramelteon as in claim 5, wherein S2 comprises S2.3, obtaining a final robust innovation sequence after iterative convergence And final weighting factor : ; ; In the formula, Is that The state estimate vector for the moment in time, And estimating the vector as the final robust state after iteration convergence.
7. The adaptive filtering method for underwater navigation based on the lower bound constraint of claim 6, wherein S3 comprises, using the recursive nature of the snow information matrix, calculating S3.1 Time prediction information matrix : ; In the formula, In the form of a state transition matrix, Is that A posterior fee-checking snow information matrix at moment; S3 comprises S3.2, calculating a measurement information matrix after non-Gaussian correction : 。
8. The adaptive filtering method for underwater navigation based on the lower bound constraint of claim 7, wherein S3 comprises S3.3, updating Time posterior Kramer lower bound : ; ; In the formula, Is that A posterior fee-checking snow information matrix at moment; Will be As a means of The time state estimates the performance lower bound of covariance.
9. The adaptive filtering method for underwater navigation based on the lower bound constraint of claim 8, wherein S4 comprises S4.1, calculating an empirical estimate of the process noise covariance at time k using modified sags Hu Sazi adaptive filtering : ; In the formula, As a forgetting factor, Is robust innovation; s4 comprises S4.2, constructing a constraint tolerance based on a lower boundary of the Keramelteon : ; In the formula, As a fault-tolerant coefficient, As the trace operator of the matrix, Is that The inverse of the prediction information matrix of the time instant.
10. The adaptive filtering method for underwater navigation based on the lower bound constraint of claim 4, wherein S4 comprises S4.3, In the time-course of which the first and second contact surfaces, ; In the time-course of which the first and second contact surfaces, , To be corrected after ; S4 comprises, S4.4, to Returning to step S2.1, the state vector at the next moment is predicted.

Description

Underwater navigation self-adaptive filtering method based on Keramelteon lower bound constraint Technical Field The invention discloses an underwater navigation self-adaptive filtering method based on a lower bound constraint of Keramelteon, and belongs to the technical field of underwater navigation positioning. Background With the increasing wide application of Autonomous Underwater Vehicles (AUV) in marine resource development and scientific investigation, fusion of underwater multisource sensors (such as LBL/INS) becomes a key technology of integrated navigation. However, complex underwater acoustic environments face two major challenges, non-gaussian heavy tail noise (e.g., outliers due to multipath effects) and time-varying characteristics of noise statistics (e.g., interference fluctuations due to sea state variations). The prior art often solves the above problems by means of robust filtering or adaptive filtering and fusion of the two. Adaptive filtering such as Sage-Husa uses a innovation sequence back-push noise covariance matrix to track noise variations, but is essentially an empirical estimate based on historical data, lacking theoretical constraints. In a non-gaussian environment, individual large amplitude outliers are very misjudged as an overall increase in system noise, resulting in covariance estimation "bloating" or even divergence, with regulatory hysteresis. Therefore, aiming at the complex coupling environment of the current underwater time-varying noise and the non-Gaussian noise, a closed-loop control mechanism which can not only utilize data to carry out self-adaptive tracking, but also utilize theoretical boundaries to prevent blind divergence is lacking. Aiming at the problem of strong coupling of non-Gaussian heavy tail noise and time-varying noise statistical characteristics in a complex underwater dynamic environment, the self-adaptive filtering method such as Sage-Husa based on innovation feedback in the prior art essentially belongs to unconstrained open-loop empirical estimation, and the updating range of a covariance matrix is monitored and limited in real time due to the lack of theoretical references based on a physical model and information quantity, so that abnormal deviation is easily misjudged as system noise increase when Jiang Ye value interference is encountered, and the technical problems of irrational expansion, parameter drift, filtering divergence and the like of noise covariance matrix estimation are caused. Disclosure of Invention The invention aims to provide an underwater navigation self-adaptive filtering method based on the lower bound constraint of the Kramer, which aims to solve the problems of irrational expansion, parameter drift and filtering divergence of noise covariance matrix estimation caused by the fact that abnormal deviation is easily misjudged as system noise increase when Jiang Ye value interference is encountered due to the fact that the updating range of a covariance matrix is monitored and limited in real time based on a theoretical reference of a physical model and information quantity in the prior art. An underwater navigation self-adaptive filtering method based on the lower bound constraint of the Keramelteon comprises the following steps: S1, establishing a state equation and an observation equation of a discrete time system, expressing measurement noise as a Gaussian scale mixed model, and expressing an edge probability density function of the measurement noise as an integral form of conditional Gaussian distribution and gamma distribution based on the Gaussian scale mixed model; S2, predicting a state vector and a state covariance matrix by using a process noise covariance matrix, setting an iteration frequency threshold at each moment, and carrying out variation iteration update, wherein the variation iteration update sequentially comprises updating an innovation covariance matrix, calculating the expectation of the square of a measurement residual error and updating a weighting factor of an auxiliary variable, and obtaining a final robust innovation sequence and a final weighting factor after reaching the iteration frequency threshold; s3, predicting an information matrix by using the Fisher information matrix, calculating a non-Gaussian corrected measurement information matrix by using a final weighting factor, and updating a posterior Kramer lower bound by using the predicted information matrix and the non-Gaussian corrected measurement information matrix; S4, calculating an empirical estimation value of a process noise covariance matrix at the previous moment by utilizing a final robust information sequence, then combining improved Sagnac Hu Sazi to adapt filtering, calculating the empirical estimation value of the process noise covariance matrix at the current moment, constructing a constraint tolerance based on a Keramer lower bound, calculating a process noise covariance matrix subjected to tolerance constraint processi