A real-time underwater acoustic positioning method based on bias perception and online priori adaptation

Abstract

A real-time underwater acoustic positioning method based on deviation perception and online prior adaptation is disclosed, relating to the fields of underwater positioning and underwater acoustic information processing. Addressing the problem of low positioning accuracy in existing underwater acoustic positioning methods under complex degradation conditions, this application explicitly incorporates beacon coordinate deviation into state variables or prior information for modeling. Combined with propagation compensation under layered sound velocity conditions, link-level robust reweighting, and an online prior adaptation mechanism for consistency adjustment, this method improves positioning accuracy under complex degradation conditions while ensuring real-time performance.

Description

Technical Field

[0001] This application relates to the field of underwater positioning and underwater acoustic information processing technology, specifically a real-time underwater acoustic positioning method based on deviation perception and online prior adaptation. Background Technology

[0002] Underwater global satellite navigation signals are difficult to propagate directly. Underwater targets, unmanned underwater vehicles, autonomous underwater robots, and marine observation platforms typically rely on underwater acoustic positioning technology for location estimation. Existing real-time underwater acoustic positioning methods commonly face problems of anomalous measurements and asynchronous degradation in underwater acoustic links during engineering applications. Affected by multipath propagation, obstruction, environmental noise, and changes in equipment status, different acoustic links may experience abnormal delays, gross errors, or short-term degradation at different times—i.e., complex degradation conditions. This leads to low positioning accuracy in existing underwater acoustic positioning methods. Summary of the Invention

[0003] The purpose of this invention is to address the problem of low positioning accuracy caused by existing underwater acoustic positioning methods under complex degradation conditions, and to provide a real-time underwater acoustic positioning method based on deviation perception and online prior adaptation.

[0004] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0005] A real-time underwater acoustic localization method based on deviation perception and online prior adaptation includes the following steps:

[0006] Step 1: Obtain one-way or equivalent one-way acoustic propagation delay observations between the target node and multiple acoustic beacons, as well as target depth, beacon depth, environmental sound speed profile information, and target motion assistance information;

[0007] Step 2: Based on target depth, beacon depth, ambient sound velocity profile information, and target motion assistance information, establish a closed mapping relationship between propagation delay and slant range or horizontal distance. Then, based on this mapping relationship and combined with one-way or equivalent one-way acoustic propagation delay observations between the target node and multiple acoustic beacons, obtain the geometric distance. ;

[0008] Step 3: Establish an augmented state model that includes the target motion state and beacon coordinate deviation, and introduce zero-sum constraints or equivalent specification constraints on the beacon coordinate deviation in the model;

[0009] Step 4: For the first The link is Measurement residuals at time Adopting Student's- Calculate link-level robust weights using a distributed or equivalent hierarchical Gaussian scale mixture model. And based on link-level robust weights Constructing effective measurement covariance ;

[0010] Step 5: Obtain the prior mean and prior covariance of the current deviation of the target, and combine them with the augmented state model obtained in Step 3, as well as the effective measurement covariance. Construct a Kalman-form filter update equation and solve it to obtain the posterior estimate of the target state and the posterior estimate of the bias at the current time, while applying zero-sum constraints.

[0011] Step 6: Based on geometric distance The target state posterior estimate and bias posterior estimate at the current moment are recursively updated to obtain the target position estimate, velocity estimate, bias estimate and its covariance, thus realizing the underwater acoustic positioning at the current moment.

[0012] Furthermore, the motion-aid information includes: velocity observation, heading observation, or inertial navigation information.

[0013] Furthermore, the specific steps for establishing the closed mapping relationship between propagation delay and slant range or horizontal distance in step 2 are as follows:

[0014] Step 21: Based on the target depth, beacon depth, ambient sound velocity profile information, and target motion assistance information, obtain the sound velocity profile intervals between the target depth and each beacon depth;

[0015] Step 22: Using the equivalent gradient approximation with equal area constraints, the sound velocity profile is approximated as a linear equivalent sound velocity model within the depth interval corresponding to each link, thereby establishing a closed mapping relationship between propagation delay and slant range or horizontal distance, specifically:

[0016] Step 221: For the first The link is used to obtain the target depth and the first link. The area of ​​the integral of the speed of sound between each beacon depth , represented as:

[0017] ,

[0018] in, Indicates the first The depth coordinates of each beacon. Represents the depth coordinates of the target. Indicates depth The sound speed function value at that location, Represents a depth integral infinitesimal element;

[0019] Step 222: Based on the integral area of ​​sound velocity An equivalent linear sound speed model is constructed to obtain a closed mapping relationship between propagation time delay and slant range or horizontal distance.

[0020] Furthermore, the augmented state vector in the augmented state model includes the target's two-dimensional or three-dimensional position, target velocity, and coordinate deviation components of multiple beacons.

[0021] Furthermore, the link-level robust weights Represented as:

[0022] ,

[0023] in, Indicates Student's- The degree of freedom parameters of the distributed or equivalent hierarchical Gaussian-scale mixture model. This represents the nominal or learned link variance parameter at a previous moment on the link.

[0024] Furthermore, the effective measurement covariance Represented as:

[0025] .

[0026] Furthermore, in step 6, the geometric distance is used as the basis for... The specific steps for recursively updating the posterior estimate and bias posterior estimate of the target state at the current moment are as follows:

[0027] Step 61: Based on geometric distance , obtain Moment-time measurement innovation vector And measurement innovation covariance matrix and based on Moment-time measurement innovation vector And measurement innovation covariance matrix Calculate the normalized squared innovation statistic. , represented as:

[0028] ,

[0029] ,

[0030] ,

[0031] in, This represents a nonlinear function from the state vector to the observation space. express State vector prediction and estimation at time t. Represents measurement function exist The first-order Jacobian matrix at that location, express Prediction and estimation of the error covariance matrix at time t. This indicates the number of active links currently participating in the update. This represents the scalar effective measurement covariance corresponding to each valid link currently involved in the update. The effective measurement covariance matrix is ​​composed in diagonal form;

[0032] Step 62: Based on Consistency adjustment factor was obtained. ;

[0033] Step 63: Link-level robust weights To obtain the link credibility index ;

[0034] Step 64: Utilize the consistency adjustment factor and link credibility index , obtain learning step size ;

[0035] Step 65: Based on learning step size Using the posterior mean, posterior covariance, and posterior second-order statistic corresponding to the posterior distribution of the deviation at the current time, the prior mean, prior covariance, and variance parameters of each link at the next time step are updated online adaptively and recursively. Specifically, this includes:

[0036] Perform an exponential smoothing update on the prior mean of the bias based on the current posterior mean of the bias;

[0037] Perform a recursive update of the bias prior covariance based on the current bias posterior covariance and the posterior second-order statistic;

[0038] The link variance parameter is updated online using a forgetting factor driven by the current measurement statistics.

[0039] Furthermore, the consistency adjustment factor Represented as:

[0040] ,

[0041] in, Indicates the consistency threshold. Indicates a stable term. This indicates the adjustment intensity parameter.

[0042] Furthermore, the link trustworthiness index Represented as:

[0043] ,

[0044] in, express The set of valid link indices that are constantly being updated in the current recursive manner.

[0045] Furthermore, the learning step size Represented as:

[0046] ,

[0047] in, Indicates the smallest learning base. This represents the learning amplitude coefficient.

[0048] The beneficial effects of this invention are:

[0049] This application explicitly incorporates beacon coordinate deviation into state variables or prior information for modeling, and combines propagation compensation under layered sound velocity conditions, link-level robust reweighting, and online prior adaptive mechanisms for consistency adjustment, thereby improving positioning accuracy under complex degradation conditions while ensuring real-time performance. Attached Figure Description

[0050] Figure 1 This is the overall flowchart of this application;

[0051] Figure 2 This is a schematic diagram of propagation compensation based on equivalent gradient under layered sound velocity profiles.

[0052] Figure 3 Schematic diagram for modeling bias-perceived augmented states;

[0053] Figure 4 This diagram illustrates link-level robust reweighting and online prior adaptive updating. Detailed Implementation

[0054] It should be noted that, where there is no conflict, the various embodiments disclosed in this application can be combined with each other.

[0055] Specific Implementation Method 1: The real-time underwater acoustic localization method based on deviation perception and online prior adaptation described in this implementation method includes the following steps:

[0056] Step 1: Obtain one-way or equivalent one-way acoustic propagation delay observations between the target node and multiple acoustic beacons, as well as target depth, beacon depth, environmental sound speed profile information, and target motion assistance information;

[0057] Step 2: Based on target depth, beacon depth, ambient sound velocity profile information, and target motion assistance information, establish a closed mapping relationship between propagation time delay and slant range or horizontal distance to obtain the geometric distance. ;

[0058] Step 3: Establish an augmented state model that includes the target motion state and beacon coordinate deviation, and introduce zero-sum constraints or equivalent specification constraints on the beacon coordinate deviation in the model;

[0059] Step 4: For the first The link at time Measurement residuals Adopting Student's- Calculate link-level robust weights using a distributed or equivalent hierarchical Gaussian scale mixture model. And based on link-level robust weights Constructing effective measurement covariance ;

[0060] Step 5: Obtain the prior mean and prior covariance of the current deviation of the target, and combine them with the augmented state model obtained in Step 3, as well as the effective measurement covariance. Construct a Kalman-form filter update equation and solve it to obtain the posterior estimate of the target state and the posterior estimate of the bias at the current time, while applying zero-sum constraints.

[0061] Step 6: Based on geometric distance The target state posterior estimate and bias posterior estimate at the current moment are recursively updated to obtain the target position estimate, velocity estimate, bias estimate and its covariance, thus realizing the underwater acoustic positioning at the current moment.

[0062] Obtaining the prior and observational information required for positioning

[0063] Acquire one-way or equivalent one-way acoustic propagation time delay observations between the target node and multiple acoustic beacons, and acquire target depth, beacon depth, environmental sound speed profile information, and target motion assistance information; wherein, the motion assistance information includes, but is not limited to, velocity observation, heading observation, inertial navigation information, or a combination thereof.

[0064] Establish a propagation compensation model under layered sound speed conditions

[0065] Based on the sound velocity profile intervals between the target depth and the depths of each beacon, an equivalent gradient approximation with equal area constraints is adopted to approximate the actual layered sound velocity profile as a linear equivalent sound velocity model within the depth interval corresponding to each link. This establishes a closed mapping relationship between propagation delay and slant range or horizontal distance, which is used to reduce the propagation mismatch error caused by the layered sound velocity profile.

[0066] Preferably, for the first The link defines the target depth and the first link. The integral area of ​​the sound velocity across each beacon depth is:

[0067] ,

[0068] An equivalent linear sound speed model is constructed within this interval, thereby obtaining a closed mapping between propagation delay and geometric distance.

[0069] Based on the one-way or equivalent one-way acoustic propagation delay observations between the target node and multiple acoustic beacons, and the closed-form mapping between propagation delay and geometric distance, the geometric distance is obtained. .

[0070] Constructing an underwater acoustic localization state model with deviation perception

[0071] An augmented state model is established, incorporating the target's motion state and beacon coordinate deviations. Preferably, the augmented state vector in the augmented state model includes the target's two-dimensional or three-dimensional position, target velocity, and coordinate deviation components of multiple beacons; more preferably, zero-sum constraints or equivalent specification constraints are introduced for all beacon coordinate deviations to eliminate the translational indistinguishability of the deviation parameters.

[0072] Build a robust measurement model and perform link-level reweighting

[0073] To address potential abnormal degradation in different acoustic links, Student's- Distribution or equivalent hierarchical Gaussian scale mixture model for the first The link at time Measurement residuals Calculate link-level robust weights:

[0074] ,

[0075] in, For Students's The degree of freedom parameters of the distributed or equivalent hierarchical Gaussian-scale mixture model. This represents the nominal or learned variance parameter at a given time step on the link. The effective measurement covariance is constructed based on the robust weights:

[0076] ,

[0077] This allows for selective suppression of abnormal links.

[0078] Perform bias-aware constraint-robust filter updates

[0079] Using the augmented state model, effective measurement covariance, and the prior mean and prior covariance of the current bias, a Kalman-form filter update equation is constructed to solve for the posterior estimate of the target state and the posterior estimate of the bias at the current time. At the same time, the zero-sum constraint (gauge constraint) is applied to ensure the identifiability and stability of the bias estimation results.

[0080] Constructing an online prior adaptive mechanism for consistency regulation

[0081] To avoid the persistent contamination of long-term biased priors by erroneous posteriors caused by abnormal measurements, an online prior adaptive mechanism based on innovation consistency and link credibility is constructed.

[0082] Preferably, the calculation time Innovation vector, innovation covariance, and normalized squared innovation statistic:

[0083] ,

[0084] ,

[0085] ,

[0086] in, This indicates the number of currently active links.

[0087] in, Indicates time The distance observation vector; It represents a nonlinear function from the state vector to the observation space; Indicates time State vector prediction estimation; Represents measurement function exist The first-order Jacobian matrix at the location; Indicates time The error covariance matrix prediction estimation; This indicates the number of active links currently participating in the update; This represents the normalized squared innovation statistic, used to measure the consistency level of the current filter update.

[0088] Furthermore, define the consistency adjustment factor:

[0089] ,

[0090] in, As a consistency threshold, As a stable term, To adjust the intensity parameters.

[0091] Furthermore, a link reliability index is constructed based on the robustness weights of each link.

[0092] ,

[0093] And define the learning step size

[0094] ,

[0095] in, As the smallest learning base, This is the learning amplitude coefficient.

[0096] Update the prior mean of the deviation, the prior covariance of the deviation, and the link hyperparameters.

[0097] Based on the obtained learning step size, the bias prior mean and bias prior covariance are updated online recursively, specifically including: ① performing exponential smoothing update on the bias prior mean; ② performing recursive update on the bias prior covariance by combining the forgetting factor with the current posterior second-order statistics; ③ using forgetting factor-driven online update on the link variance hyperparameter.

[0098] When the consistency or reliability of the current epoch is poor, the learning step size is reduced or even set to zero, thereby blocking the transmission of unreliable posterior to prior; when the statistics of the current epoch are consistent and the link quality is good, the biased prior is allowed to gradually absorb reliable posterior information.

[0099] Output real-time positioning results

[0100] The system outputs target position estimates, velocity estimates, bias estimates, and their covariance, thus achieving underwater acoustic localization at the current moment.

[0101] Optionally, fixed-hysteresis smoothing can be performed on the most recent few epochs to further improve trajectory smoothness and estimation accuracy.

[0102] This application first obtains acoustic propagation delay observations between the target to be located and multiple acoustic beacons, target depth, depths of each acoustic beacon, environmental sound speed profiles, and target motion assistance information. The target motion assistance information may be velocity observations, heading observations, inertial navigation information, or a combination thereof.

[0103] Then, based on the environmental sound speed profile information, the acoustic propagation process between the target to be located and each acoustic beacon is compensated to obtain the distance constraint information corresponding to each acoustic link.

[0104] Subsequently, an augmented state model incorporating the target motion state and acoustic beacon coordinate deviation is constructed, and constraints are applied to the acoustic beacon coordinate deviation to obtain a deviation-aware localization model.

[0105] At each positioning epoch, robust reweighting is performed on each acoustic link based on the current predicted state and measurement residual to construct the link-level effective measurement covariance. Then, based on the augmented state model, the distance constraint information, the link-level effective measurement covariance, and the current bias prior information, a constraint robust filtering update is performed to obtain the target state posterior estimate and bias posterior estimate at the current moment.

[0106] Furthermore, based on the current innovation consistency index and link credibility index, an online prior adaptive learning step size is constructed. Based on this learning step size, the bias prior mean, bias prior covariance, and link variance parameters are updated using a forgetting-based one-step recursive update. When the innovation consistency index does not meet preset conditions, the bias prior update is reduced or stopped to avoid unreliable posteriors continuously polluting long-term bias priors.

[0107] Finally, the real-time positioning result of the target to be located is output.

[0108] At the depth of the target to be located and the first Within the depth range between acoustic beacon depths, an equivalent gradient approximation model is established based on the environmental sound velocity profile. Preferably, within the depth range, an equivalent linear sound velocity model is constructed through interval equal area constraints, so that the equivalent model approximates the real layered sound velocity profile in an integral sense.

[0109] Based on this, the acoustic propagation delay is mapped to slant range or horizontal distance according to the equivalent linear sound speed model, thereby constructing the distance constraint information required for positioning.

[0110] The augmented state model includes the target position state, the target velocity state, and the coordinate deviation states of multiple acoustic beacons. The acoustic beacon coordinate deviation is used to characterize the unknown offset of the actual beacon coordinates relative to the nominal coordinates.

[0111] To eliminate the unidentifiable nature of the deviation parameters, constraints are imposed on the coordinate deviations of multiple acoustic beacons. Preferably, the constraints are zero-sum constraints; in other embodiments, normative constraints or equivalent constraints may also be used.

[0112] For each acoustic link, the robust weight is calculated using a robust reweighting method based on the Student's-t distribution, given the measurement residual at the current time. The robust weight is determined by the measurement residual, the degree of freedom parameter, and the variance parameter of the corresponding link.

[0113] Based on the obtained link robust weights and the variance parameters corresponding to each link, the effective measurement covariance of each acoustic link is constructed. Then, the effective measurement covariances of each link are combined into the link-level effective measurement covariance matrix at the current time, which is used for subsequent constraint robust filtering updates.

[0114] For each acoustic link, the robust weight is calculated using a robust reweighting method based on the Student's-t distribution, given the measurement residual at the current time. The robust weight is determined by the measurement residual, the degree of freedom parameter, and the variance parameter of the corresponding link.

[0115] Based on the obtained link robust weights and the variance parameters corresponding to each link, the effective measurement covariance of each acoustic link is constructed. Then, the effective measurement covariances of each link are combined into the link-level effective measurement covariance matrix at the current time, which is used for subsequent constraint robust filtering updates.

[0116] At the current positioning epoch, a joint estimation model of the target state and the deviation state is constructed based on the current predicted state, the current link-level effective measurement covariance, the current deviation prior mean, and the deviation prior covariance. Subsequently, constrained robust filtering is performed to update the model while satisfying the deviation constraints, resulting in the posterior estimates of the target state and the deviation at the current time.

[0117] The innovation consistency index is constructed based on the innovation vector and innovation covariance at the current moment; the link credibility index is constructed based on the robust weights corresponding to the currently effective acoustic links. The online prior adaptive learning step size is jointly determined by the innovation consistency index and the link credibility index.

[0118] When the innovation consistency index exceeds the preset threshold, the online prior adaptive learning step size is set to zero or reduced to the preset range to suppress the transmission of unreliable posterior to prior. When the innovation consistency index does not exceed the preset threshold, the online prior adaptive learning step size is adjusted according to the link credibility index so that statistically reliable posterior information can be gradually incorporated into long-term biased priors.

[0119] The prior mean of the bias is updated recursively using exponential smoothing; the prior covariance of the bias is updated recursively based on the forgetting factor and the current posterior statistic of the bias; the link variance parameter is updated online based on the forgetting factor. This update method is a one-step recursive approach, eliminating the need for inner fixed-point iterations or multiple rounds of variational coordinate ascent.

[0120] When the innovation consistency index does not meet the preset conditions, the bias prior mean and bias prior covariance remain unchanged, or only updates below the preset threshold are performed, thereby suppressing the erroneous prior contraction caused by abnormal epochs.

[0121] After outputting the real-time positioning results, fixed-hysteresis smoothing can be performed on the most recent few epochs to further improve trajectory smoothness and positioning accuracy. The fixed-hysteresis smoothing can be based on the most recent... The posterior result of each epoch is executed.

[0122] A real-time underwater acoustic positioning system based on deviation sensing and online prior adaptation includes:

[0123] The observation and acquisition module is used to acquire acoustic propagation delay observations between the target to be located and multiple acoustic beacons, target depth, depth of each acoustic beacon, environmental sound speed profile information, and target motion assistance information;

[0124] The propagation compensation module is used to compensate for the acoustic propagation process based on the environmental sound speed profile information, and to obtain the distance constraint information corresponding to each acoustic link;

[0125] The state modeling module is used to construct an augmented state model that includes the target motion state and acoustic beacon coordinate deviation, and to impose constraints on the acoustic beacon coordinate deviation.

[0126] The robust reweighting module is used to robustly reweight each acoustic link based on the current prediction state and measurement residuals and construct the link-level effective measurement covariance.

[0127] The constraint filtering module is used to perform constraint robust filtering updates;

[0128] The online prior adaptive module is used to construct the online prior adaptive learning step size based on the innovation consistency index and the link credibility index, and to perform forgetting-style one-step recursive updates on the bias prior mean, bias prior covariance and link variance parameters.

[0129] The results output module is used to output the real-time positioning results of the target to be located.

[0130] This embodiment provides an electronic device and a computer-readable storage medium.

[0131] The electronic device includes a memory, a processor, and a computer program stored in the memory. When the processor executes the computer program, it implements the real-time underwater acoustic positioning method described in any of the foregoing embodiments.

[0132] The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the real-time underwater acoustic positioning method described in any of the foregoing embodiments.

[0133] In this embodiment, the link-level robust reweighting, effective measurement covariance construction, consistency adjustment factor, online prior adaptive learning step size, and bias prior update process in the method of this application are further explained.

[0134] Let the first Acoustic links at time The measurement residual is The variance parameter of the corresponding link is The degree of freedom parameters are Based on Student's Robust modeling methods, the first The robust weight of an acoustic link can be expressed as:

[0135] ,

[0136] Based on the robust weights, construct the first... Effective measurement covariance of an acoustic link:

[0137] ,

[0138] Furthermore, by combining the effective measurement covariances of each acoustic link, the link-level effective measurement covariance matrix at the current moment is obtained:

[0139] ,

[0140] Let the innovation vector at the current moment be... Innovation covariance is The current number of valid links is Then we can construct the normalized squared innovation statistic:

[0141] ,

[0142] Based on the normalized squared innovation statistic, a consistency adjustment factor is constructed:

[0143] ,

[0144] in, As a consistency threshold, For numerically stable terms, To adjust the intensity parameters.

[0145] Furthermore, based on the robust weights of the currently effective acoustic links, a link reliability index is constructed:

[0146] ,

[0147] in, Indicates time The set of valid links.

[0148] Based on this, define the online prior adaptive learning step size:

[0149] ,

[0150] in, As the smallest learning base, This represents the learning amplitude coefficient. Therefore, it can be seen that when the consistency of the current epoch is poor, the consistency adjustment factor... The learning step size is increased when the current epoch has good consistency and the link has high reliability, thereby enhancing the transmission of reliable posterior information to the prior.

[0151] Set time The biased posterior estimate is The posterior covariance of the bias is The bias prior mean is The prior covariance of the bias is When the update conditions are met at the current epoch, the prior mean of the deviation can be updated recursively in the following manner:

[0152] ,

[0153] The bias prior covariance can be updated recursively in the following manner:

[0154] ,

[0155] When the innovation consistency index does not meet the preset conditions, the prior mean of the deviation is maintained. and the aforementioned bias prior covariance The system either remains unchanged or performs only minor updates below a preset threshold, thereby suppressing erroneous prior contraction caused by abnormal epochs and the continuous propagation of unreliable posterior information to long-term priors.

[0156] In a preferred embodiment, to enable online updates of the link variance parameters, a forgetting statistic is introduced for each link. and And it is recursively applied as follows:

[0157] ,

[0158] in, It is a forgetting factor.

[0159] When the current epoch meets the update conditions, further execution is performed on the valid links:

[0160] ,

[0161] And obtain the updated link variance parameters:

[0162] ,

[0163] in, It is a stable term.

[0164] In another implementation, when the innovation consistency index of the current epoch does not meet the preset conditions, the link variance parameter is maintained. The forgetting statistic remains unchanged, or is stopped. and Effective incremental updates are used to avoid abnormal links causing continuous contamination of subsequent epoch variance estimates.

[0165] It exhibits strong explicit deviation perception capability. This application explicitly incorporates beacon coordinate deviation into the positioning modeling process, which can significantly reduce the contamination of target position estimation by geometric mismatch error.

[0166] This invention is applicable to stratified marine environments. It introduces a propagation compensation mechanism based on equivalent gradients, which can effectively reduce the propagation delay and range mapping mismatch caused by stratified sound velocity profiles, thereby improving positioning accuracy in complex marine environments.

[0167] This application is robust to anomalous measurements. It utilizes a link-level Student's- The reweighting mechanism selectively suppresses asynchronous degraded acoustic links, thereby enhancing stability under gross errors, sudden anomalies, and local link degradation conditions.

[0168] Online prior adaptive mechanisms can suppress the propagation of erroneous priors. This application employs an online prior adaptive strategy with consistency adjustment, allowing reliable posterior information to accumulate gradually, while anomalous epochs do not continuously contaminate long-term biased priors, thus improving accuracy and consistency over long-term operation.

[0169] With low computational complexity, it is suitable for real-time implementation. This application constructs the bias prior update as an explicit one-step recursion, avoiding inner variational coordinate ascent and fixed-point iteration, thus reducing computational complexity while maintaining robustness, making it suitable for embedded and real-time platform deployment.

[0170] It has strong engineering adaptability. This application can simultaneously handle propagation mismatch, beacon bias, abnormal measurements, and measurement loss, and is applicable to scenarios such as seabed array positioning, underwater unmanned platform navigation, underwater target tracking, and marine observation network positioning.

[0171] It should be noted that the specific embodiments are merely explanations and illustrations of the technical solution of the present invention and should not be used to limit the scope of protection. Any modifications made in accordance with the claims and specification of the present invention that are only partial should still fall within the protection scope of the present invention.

Claims

1. A real-time underwater acoustic localization method based on deviation perception and online prior adaptation, characterized in that... Includes the following steps: Step 1: Obtain one-way or equivalent one-way acoustic propagation delay observations between the target node and multiple acoustic beacons, as well as target depth, beacon depth, environmental sound speed profile information, and target motion assistance information; Step 2: Based on target depth, beacon depth, ambient sound velocity profile information, and target motion assistance information, establish a closed mapping relationship between propagation delay and slant range or horizontal distance. Then, based on this mapping relationship and combined with one-way or equivalent one-way acoustic propagation delay observations between the target node and multiple acoustic beacons, obtain the geometric distance. ; Step 3: Establish an augmented state model that includes the target motion state and beacon coordinate deviation, and introduce zero-sum constraints or equivalent specification constraints on the beacon coordinate deviation in the model; Step 4: For the first The link is Measurement residuals at time Adopting Student's- Calculate link-level robust weights using a distributed or equivalent hierarchical Gaussian scale mixture model. And based on link-level robust weights Constructing effective measurement covariance ; Step 5: Obtain the prior mean and prior covariance of the current deviation of the target, and combine them with the augmented state model obtained in Step 3, as well as the effective measurement covariance. Construct a Kalman-form filter update equation and solve it to obtain the posterior estimate of the target state and the posterior estimate of the bias at the current time, while applying zero-sum constraints. Step 6: Based on geometric distance The target state posterior estimate and bias posterior estimate at the current moment are recursively updated to obtain the target position estimate, velocity estimate, bias estimate and its covariance, thus realizing the underwater acoustic positioning at the current moment.

2. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 1, characterized in that... The motion-aided information includes: velocity observation, heading observation, or inertial navigation information.

3. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 1, characterized in that... The specific steps for establishing the closed mapping relationship between propagation delay and slant range or horizontal distance in step 2 are as follows: Step 21: Based on the target depth, beacon depth, ambient sound velocity profile information, and target motion assistance information, obtain the sound velocity profile intervals between the target depth and each beacon depth; Step 22: Using the equivalent gradient approximation with equal area constraints, the sound velocity profile is approximated as a linear equivalent sound velocity model within the depth interval corresponding to each link, thereby establishing a closed mapping relationship between propagation delay and slant range or horizontal distance, specifically: Step 221: For the first The link is used to obtain the target depth and the first link. The area of ​​the sound velocity integral between each beacon depth , is represented as: ， in, Indicates the first The depth coordinates of each beacon. Represents the depth coordinates of the target. Indicates depth The sound speed function value at that location, Represents a depth integral infinitesimal element; Step 222: Based on the integral area of ​​sound velocity An equivalent linear sound speed model is constructed to obtain a closed mapping relationship between propagation time delay and slant range or horizontal distance.

4. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 3, characterized in that... The augmented state vector in the augmented state model includes the target's two-dimensional or three-dimensional position, target velocity, and coordinate deviation components of multiple beacons.

5. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 4, characterized in that... The link-level robust weight Represented as: ， in, Indicates Student's- The degree of freedom parameters of the distributed or equivalent hierarchical Gaussian-scale mixture model. This represents the nominal or learned link variance parameter at a previous moment on the link.

6. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 5, characterized in that... The effective measurement covariance Represented as: 。 7. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 6, characterized in that... Step 6 is based on geometric distance The specific steps for recursively updating the posterior estimate and bias posterior estimate of the target state at the current moment are as follows: Step 61: Based on geometric distance , obtain Moment-time measurement innovation vector And measurement innovation covariance matrix and based on Moment-time measurement innovation vector And measurement innovation covariance matrix Calculate the normalized squared innovation statistic. , is represented as: ， ， ， in, This represents a nonlinear function from the state vector to the observation space. express State vector prediction and estimation at time t. Represents measurement function exist The first-order Jacobian matrix at that location, express Prediction and estimation of the error covariance matrix at time t. This indicates the number of active links currently participating in the update. This represents the scalar effective measurement covariance corresponding to each valid link currently involved in the update. The effective measurement covariance matrix is ​​composed in diagonal form; Step 62: Based on Consistency adjustment factor was obtained. ; Step 63: Link-level robust weights To obtain the link credibility index ; Step 64: Utilize the consistency adjustment factor and link credibility index , obtain learning step size ; Step 65: Based on learning step size Using the posterior mean, posterior covariance, and posterior second-order statistic corresponding to the posterior distribution of the deviation at the current time, the prior mean, prior covariance, and variance parameters of each link at the next time step are updated online adaptively and recursively. Specifically, this includes: Perform an exponential smoothing update on the prior mean of the bias based on the current posterior mean of the bias; Perform a recursive update of the bias prior covariance based on the current bias posterior covariance and the posterior second-order statistic; Perform online updates of the link variance parameters based on the forgetting factor driven by the current measurement statistics.

8. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 7, characterized in that... The consistency adjustment factor Represented as: ， in, Indicates the consistency threshold. Indicates a stable term. This indicates the adjustment intensity parameter.

9. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 8, characterized in that... The link reliability index Represented as: ， in, express The set of valid link indices that are constantly being updated in the current recursive manner.

10. The real-time underwater acoustic localization method based on deviation perception and online prior adaptation according to claim 9, characterized in that... The learning step size Represented as: ， in, Indicates the smallest learning base. This represents the learning amplitude coefficient.

Images