In-situ correction control method for deep-sea magnetometer
By combining reinforcement learning agents and Kalman filtering, dynamic maneuver commands are generated, solving the problem of deep-sea magnetometer calibration and achieving high-precision deep-sea magnetometer calibration. This adapts to different carriers and mission requirements and avoids hardware installation difficulties.
Patent Information
- Application Number
- CN202511925898.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies are insufficient for efficiently, conveniently, and accurately correcting the carrier magnetic deviation of deep-sea magnetometers in deep-sea environments. Traditional methods rely on the maneuvering navigation mode of surface vessels, which cannot achieve high-precision correction on deep-sea platforms, and the hardware installation is difficult.
By combining reinforcement learning agents and state-extended strong tracking Kalman filters, and using the carrier's built-in sensors and global geomagnetic model, maneuver commands for small-angle attitude adjustment and micro-depth change are dynamically generated. An interference magnetic field model is constructed and corrected, and high-precision correction is achieved by refining the parameters using L1 regularized least squares.
Achieving a calibration accuracy better than 5nT in deep-sea environments, without the need for manual intervention or buoyancy calibration, adapting to different carriers and mission requirements, solving the hardware installation problem in deep-sea high-pressure sealed environments, and improving calibration efficiency and accuracy.
Smart Images

Figure CN121657148A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of magnetometer technology, and in particular to an in-situ calibration and control method for a deep-sea magnetometer. Background Technology
[0002] Magnetometers, especially high-precision marine magnetometers, are key equipment for deep-sea geophysical exploration, underwater target detection, and marine scientific research. Their core function is to accurately measure the geomagnetic field vector information. However, similar to magnetic compasses on ships, the measurement accuracy of magnetometers operating in the deep-sea environment is significantly affected by the magnetic field of the carrier (such as a deep-sea submersible, seabed observation station, or towed body) itself (i.e., "ship magnetism" or more broadly, "carrier magnetism"). This interference causes the actual measurement value of the magnetometer to deviate from the true geomagnetic field value, producing a systematic error, which can be called "magnetometer carrier magnetic deviation" or, by analogy, "magnetometer self-error".
[0003] The traditional principle of eliminating carrier magnetic interference is based on the self-correction of magnetic compasses: that is, an additional magnetic field with the same properties, equal magnitude, and opposite direction as the carrier magnetic field is introduced to counteract or significantly reduce the influence of the carrier magnetic field on the sensor, thereby creating a measurement environment around the sensor probe that is as close as possible to a pure geomagnetic field, ensuring that the magnetometer can accurately sense the true direction and intensity of the geomagnetic field.
[0004] Currently, the mainstream methods widely used for magnetic compass correction on surface vessels are the Airey's Method and the Flinders-Bar Method. The core operational procedures of these methods require the vessel to maneuver strictly according to multiple prescribed headings (usually eight main compass headings) in open, designated waters. By observing the deviation of the magnetic compass from a reference (such as a gyrocompass or astronomical bearing) on different headings, corresponding hard magnets (magnetic rods) and soft iron (Follower iron, etc.) correctors are calculated and applied to compensate for the deviations on each heading. Although the Airey's Method has become the industry standard due to its relative ease of operation, the successful implementation of both the Airey's Method and the Flinders-Bar Method highly depends on this specific maneuvering pattern of the vessel on the water.
[0005] However, directly applying this surface magnetic compass correction mode to deep-sea magnetometers faces fundamental challenges and limitations:
[0006] 1. Deep-sea operation platforms (such as AUVs, ROVs, deep-sea submersible platforms, towed bodies, etc.) typically lack or find it difficult to perform the same flexible, rapid, and precise multi-directional maneuvering underwater as surface vessels. The deep-sea environment is complex, and precisely controlling course and position is inherently difficult, energy-intensive, and poses safety risks.
[0007] 2. The deep-sea environment (especially the deep seabed far from the surface) does not have the conditions of "open water". The seabed topography, nearby magnetic objects, ocean currents, etc. may introduce additional interference, making it almost impossible to replicate the surface correction environment in situ in the deep sea.
[0008] 3. Traditional hard magnet and soft iron calibrators are primarily designed for magnetic compasses, and their installation location, adjustment methods, and adaptability to higher-precision magnetometers (such as optical pumps, Overhauser, and SQUID magnetometers) may be insufficient. In the high-pressure, sealed environment of the deep sea, installing and adjusting these physical calibrators is extremely difficult or even infeasible.
[0009] 4. Requiring deep-sea platforms to frequently surface or surface in specific areas for calibration will greatly reduce operational efficiency, increase costs, and may interrupt important in-situ observation missions.
[0010] In summary, the application of existing magnetic compass calibration methods based on surface maneuvering to in-situ calibration of deep-sea magnetometers aims to meet the needs for efficient, convenient, and high-precision calibration at the operational site (in-situ), and we look forward to a satisfactory solution! Summary of the Invention
[0011] To address the aforementioned technical problems, the present invention provides an in-situ calibration and control method for a deep-sea magnetometer, the method comprising the following steps:
[0012] S01. The original magnetic field measurement value is acquired in real time through the magnetometer, inertial navigation unit, and depth sensor mounted on the carrier. Carrier attitude angle angular velocity ,depth Based on the carrier location and the global geomagnetic model, the geomagnetic reference field was calculated. and its rate of change ;
[0013] S02, Construct the covariance matrix of the interference model parameters. Using the reinforcement learning agent as input, the system dynamically generates yaw angle increments while satisfying the maneuver constraints of the deep-sea operation platform. Pitch angle increment Incremental depth and stabilization time The sequence of maneuver commands;
[0014] S03. Establish a carrier interference magnetic field model. ,in, This is the hard magnetic interference vector. It is a soft magnetic induction matrix. For eddy current disturbance values, state-extended strong tracking Kalman filtering is used to jointly estimate model parameters. With background magnetic field And refine the parameters using least squares with L1 regularization;
[0015] S04, Based on interference parameters Calculate the interfering magnetic field and from Subtract the mid-vector to output the corrected geomagnetic field. .
[0016] Preferably, the reward function of the reinforcement learning agent in step S02 Designed as follows:
[0017] ;
[0018] in, The information entropy reward weight, Let be the parameter information entropy at time t. The parameter information entropy at time t-1, For the maximum permissible yaw angle, The maximum permissible pitch angle, For the maximum allowable depth, As the weight of maneuver cost, It is a negative exponential decay term for energy consumption. Energy consumption reward weight.
[0019] Preferably, the step S03 of obtaining the regularized least squares refinement parameter includes the following steps:
[0020] S31, State Extension Modeling: The background magnetic field... With interference parameters Jointly construct state vector The state equation is defined as ,in: This is the projection of the actual geomagnetic field onto the carrier coordinate system. Let be the angular velocity vector of the carrier at time k. Let k be the joint state vector at time k. Let k be the joint state vector when k-1.
[0021] S32, Gradual Extinction Factor Adaptive Adjustment: ,in, For the innovation vector, To observe the Jacobian matrix, To predict the covariance matrix, R is the error value between the magnetometer and the pose sensor;
[0022] S33, Covariance Reset: This enhances the ability to track parameter mutations, including: The process noise covariance matrix is... It is the variance matrix;
[0023] S34. Constructing incremental equations: Where Z is the observed residual value, To increment the parameter The partial derivatives;
[0024] S35, Solve the optimization problem with L1 constraint as follows: ,in, , The sum of the absolute values of the parameter increments.
[0025] Preferably, the eddy current interference value The discretization implementation is as follows:
[0026] ;
[0027] in, The value of eddy current disturbance at time k-1. Let be the eddy current disturbance value at time k.
[0028] Preferably, steps S3 and S4 are implemented using a dual-threaded architecture, specifically including:
[0029] Background thread estimation: Execution cycle T est ≥1s, execute STEKF-RLS algorithm update ;
[0030] Foreground compensation thread: Execution cycle T c ≤1ms, based on update Real-time computing and output ;
[0031] Inter-thread data transfer: achieved through a circular buffer. The asynchronous update has a buffer depth of 1 to ensure that the compensation thread always gets the latest parameters without lock contention.
[0032] Preferably, the geomagnetic reference field in step S01 The calculation includes dynamic error correction:
[0033] ;
[0034] in, The static geomagnetic field calculated by the WMM global model. This is a real-time diurnal variation correction based on interpolation of data from the International Geomagnetic Network. This is the error feedback value.
[0035] Preferably, the conditions for terminating the above-mentioned output corrected geomagnetic field in step S04 include:
[0036] S41, Standard deviation of hard magnetic component ;
[0037] Residual magnetic field fluctuations after S42's five consecutive maneuvers .
[0038] The present invention has at least the following beneficial effects:
[0039] 1. By fusing reinforcement learning intelligent decision-making with the STEKF-RLS estimation algorithm, a correction accuracy better than 5 nT is achieved in the heavily constrained environment of the deep sea. Traditional methods (such as the Ally method) rely on complete rotational maneuvers, while this invention only requires small-angle attitude adjustments. ) and micro-depth changes ( By dynamically optimizing the maneuver path to maximize information entropy, the efficiency of parameter estimation under limited data is significantly improved.
[0040] 2. Dynamic anti-interference is achieved through a time-varying eddy current model and L1 regularized parameter estimation. The eddy current model explicitly incorporates the carrier angular velocity. The cross product term accurately characterizes the changes in the induced magnetic field caused by rotation, while L1 regularization suppresses overfitting through sparse constraints and can still maintain stability when data stimulation is insufficient.
[0041] 3. No manual intervention or buoy calibration is required. The intelligent agent learns multi-scenario strategies through simulation pre-training. In the online phase, it only needs to receive sensor data to dynamically generate maneuver commands, adapting to different carriers and mission requirements, and completely solving the hardware installation problem in the high-pressure sealed environment of the deep sea. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 The flowchart illustrates an in-situ calibration control method for a deep-sea magnetometer, as provided in Embodiment 1 of the present invention. Detailed Implementation
[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0045] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0046] Example 1
[0047] This embodiment provides an in-situ calibration control method for a deep-sea magnetometer, the method including the following steps, such as... Figure 1 As shown:
[0048] S01. The original magnetic field measurement value is acquired in real time through the magnetometer, inertial navigation unit, and depth sensor mounted on the carrier. Carrier attitude angle angular velocity ,depth Based on the carrier location and the global geomagnetic model, the geomagnetic reference field was calculated. and its rate of change ;
[0049] Specifically, the geomagnetic reference field The calculation includes dynamic error correction:
[0050] ;
[0051] in, The static geomagnetic field calculated by the WMM global model. This is a real-time diurnal variation correction based on interpolation of data from the International Geomagnetic Network. This is the error feedback value.
[0052] In the aforementioned technologies, the magnetometer retrieves raw magnetic field data. Gyroscope: records the carrier's attitude (roll) , looking up ,course ) and rotation speed (angular velocity) Depth gauge: used to record the current diving depth Its design aims to avoid collisions with the seabed or surfacing. Traditional methods suffer from data asynchrony, leading to distortion of the interference model (e.g., the magnetic field only responds one second after an attitude change). The above steps, however, utilize only the carrier's built-in sensors and a common geomagnetic data source to achieve fully autonomous in-situ reference field generation. Geomagnetic variation rate With the angular velocity of the carrier Linked modeling solves the problem of "difficulty in observing slow-moving eddies".
[0053] S02, Construct the covariance matrix of the interference model parameters. Using the reinforcement learning agent as input, the system dynamically generates yaw angle increments while satisfying the maneuver constraints of the deep-sea operation platform. Pitch angle increment Incremental depth and stabilization time The sequence of maneuver commands, in which, , , , .
[0054] Specifically, the reward function of a reinforcement learning agent. Designed as follows:
[0055] ;
[0056] in, The information entropy reward weight, Let be the parameter information entropy at time t. The parameter information entropy at time t-1, For the maximum permissible yaw angle, The maximum permissible pitch angle, For the maximum allowable depth, As the weight of maneuver cost, It is a negative exponential decay term for energy consumption. Energy consumption reward weight.
[0057] Among the aforementioned technologies, in deep-sea geophysical exploration, magnetic field interference of the carrier is the core challenge restricting the accuracy of magnetometers. Traditional correction methods (such as the Ally method) require the carrier to perform a complete eight-directional rotational maneuver, but deep-sea AUVs / ROVs are constrained by complex ocean currents, terrain obstacles, and energy limitations, making such maneuvers difficult to achieve. How to obtain the maximum amount of parameter information with minimal action cost has become the key to breaking through in-situ deep-sea correction. By employing a reinforcement learning agent for decision-making, the information is as follows: Input state: parameter covariance matrix. (Dimension: nT²), its determinant value Reflecting the overall uncertainty of the model; real-time attitude angle of the carrier Compared to historical maneuver sequences (to avoid repeating actions). Hard constraints on the maneuver space: Yaw angle increment: To prevent instability and pitch angle increments caused by sudden bow changes: To avoid exceeding the pitch angle limit and increasing the depth: To adapt to seabed topography and stabilization time: To balance data quality and energy consumption, a reward function that maximizes information entropy is calculated. In the above formula, , Essentially, it's a measure of information entropy related to parameter uncertainty. This factor drives the agent to choose a method that increases the "volume" of the covariance matrix. The action that minimizes the loss is the action that maximizes the trace of the Fisher information matrix. To normalize the maneuver penalty, the denominator 15° / 10° / 5m represents the upper limit of the motion space, making the costs of actions in different dimensions comparable. For exponentially decaying energy consumption constraints, this term approaches 0 when the cumulative energy consumption nears the platform threshold, forcing the cessation of inefficient maneuvers. Through information entropy-driven reinforcement learning, the "optimal maneuver step" was found in the highly constrained deep-sea battlefield. This not only solves the engineering challenge of magnetometer calibration but also provides a universal framework for the "perception-decision" closed loop of autonomous underwater agents.
[0058] S03. Establish a carrier interference magnetic field model. ,in, This is the hard magnetic interference vector. It is a soft magnetic induction matrix. For eddy current disturbance values, state-extended strong tracking Kalman filtering is used to jointly estimate model parameters. With background magnetic field And refine the parameters through regularized least squares;
[0059] Specifically, obtaining the L1 regularized least squares refinement parameters includes the following steps:
[0060] S31, State Extension Modeling: Incorporating the Background Magnetic Field With interference parameters Jointly construct state vector The state equation is defined as ,in: This is the projection of the actual geomagnetic field onto the carrier coordinate system. Let be the angular velocity vector of the carrier at time k. Let k be the joint state vector at time k. Let k be the joint state vector when k-1.
[0061] S32, Gradual Extinction Factor Adaptive Adjustment: ,in, For the innovation vector, To observe the Jacobian matrix, To predict the covariance matrix, R is the error value between the magnetometer and the pose sensor;
[0062] S33, Covariance Reset: This enhances the ability to track parameter mutations, including: The process noise covariance matrix is... It is the variance matrix;
[0063] S34. Constructing incremental equations: Where Z is the observed residual value, To increment the parameter The partial derivatives;
[0064] S35, Solve the optimization problem with L1 constraint as follows: ,in, , The sum of the absolute values of the parameter increments.
[0065] Furthermore, eddy current interference value The discretization implementation is as follows:
[0066] ;
[0067] in, The value of eddy current disturbance at time k-1. Let be the eddy current disturbance value at time k.
[0068] The above is achieved through explicit parameterization. and The hard magnetic component estimation error. During carrier variable-speed maneuvers, the discrete model reduces the eddy current interference residual from >10nT in the traditional model to <3nT. The state extension design eliminates geomagnetic reference map errors, making... The estimation accuracy is improved to <5nT. The fading factor resists divergence under low-excitation data (e.g., the carrier only yaws by 5∘), with STEKF suppressing parameter drift by 60%. L1 regularization resists overfitting: when maneuver data is insufficient, sparse constraints will reduce the soft magnetic matrix... Irrelevant terms are compressed to near zero to avoid ill-conditioned models. Through a three-level technology chain of "physical mechanism modeling → state extension estimation → regularization refinement", "high-precision interference separation" and "anti-divergence parameter convergence" are achieved in the deep-sea scenario with strong noise and low excitation, laying the core foundation for all-digital compensation (step S04) and directly contributing to the terminal performance with a correction accuracy of ≤5nT.
[0069] S04, Based on interference parameters Calculate the interfering magnetic field and from Subtract the mid-vector to output the corrected geomagnetic field. .
[0070] It should be noted that steps S3 and S4 above are implemented using a dual-thread architecture, specifically including:
[0071] Background thread estimation: Execution cycle T est ≥1s, execute STEKF-RLS algorithm update ;
[0072] Foreground compensation thread: Execution cycle T c ≤1ms, based on update Real-time computing and output ;
[0073] Inter-thread data transfer: achieved through a circular buffer. The asynchronous update has a buffer depth of 1 to ensure that the compensation thread always gets the latest parameters without lock contention.
[0074] Furthermore, the condition for terminating the above-mentioned corrected geomagnetic field output in step S04 is as follows:
[0075] S41, Standard deviation of hard magnetic component ;
[0076] Residual magnetic field fluctuations after S42's five consecutive maneuvers .
[0077] The calibration will stop only when both steps S41 and S42 are satisfied; otherwise, the calibration operation will continue.
[0078] Based on the above, and in conjunction with steps S03 to S04, this step utilizes the real-time pose of the carrier. Geomagnetic reference field and its rate of change (Depend on (Approximate), combined with estimated parameters Simultaneous calculation of three types of interference components: hard magnetic interference: constant vector ( (subvectors), soft magnetic interference: matrix multiplication ( Depend on Reconstructed into a 3×3 matrix), eddy current interference: Discrete equations The above soft magnetic interference calculation uses a pre-calculated matrix lookup table method (e.g., soft magnetic terms). The nine scalar multiply-accumulate operations are parallelized to ensure that the time taken for a single compensation calculation is ≤10μs. Before compensation, and Unify to the carrier coordinate system and eliminate coordinate system deviation through the attitude matrix (provided by INS).
[0079] Furthermore, in the aforementioned techniques, vector subtraction directly cancels the superposition effect of the carrier's magnetic field on geomagnetic measurements, reducing the hard / soft magnetic coupling error. Eddy current model discretization accurately matches the carrier's slow rotation (angular velocity). This reduces the interference delay and avoids the 10-20 nT residual fluctuations caused by neglecting eddy currents in traditional methods.
[0080] This embodiment achieves a correction accuracy better than 5 nT in a heavily constrained deep-sea environment by combining reinforcement learning intelligent decision-making with the STEKF-RLS estimation algorithm. Traditional methods (such as the Ally method) rely on complete rotational maneuvers, while this invention only requires small-angle attitude adjustments. ) and micro-depth changes ( First, by dynamically optimizing the maneuver path to maximize information entropy, the efficiency of parameter estimation under limited data is significantly improved. Second, dynamic anti-interference is achieved through time-varying eddy current model and L1 regularized parameter estimation. The eddy current model explicitly incorporates the carrier angular velocity. The cross product term accurately characterizes the changes in the induced magnetic field caused by rotation, while L1 regularization suppresses overfitting through sparse constraints, maintaining stability even with insufficient data stimulation. Furthermore, no manual intervention or buoyancy calibration is required. The agent masters multi-scenario strategies through simulation pre-training, and in the online phase, it only needs to receive sensor data to dynamically generate maneuver commands, adapting to different carriers and mission requirements, completely solving the hardware installation challenges in the high-pressure, sealed environment of the deep sea.
[0081] Example 2
[0082] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps:
[0083] The original magnetic field measurements are acquired in real time by a magnetometer, inertial navigation unit, and depth sensor mounted on the carrier. Carrier attitude angle angular velocity ,depth Based on the carrier location and the global geomagnetic model, the geomagnetic reference field was calculated. and its rate of change ;
[0084] Construct the covariance matrix of the interference model parameters Using the reinforcement learning agent as input, the system dynamically generates yaw angle increments while satisfying the maneuver constraints of the deep-sea operation platform. Pitch angle increment Incremental depth and stabilization time The sequence of maneuver commands;
[0085] Establish a carrier interference magnetic field model ,in, This is the hard magnetic interference vector. It is a soft magnetic induction matrix. For eddy current disturbance values, state-extended strong tracking Kalman filtering is used to jointly estimate model parameters. With background magnetic field And refine the parameters using least squares with L1 regularization;
[0086] Based on interference parameters Calculate the interfering magnetic field and from Subtract the mid-vector to output the corrected geomagnetic field. .
[0087] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0089] Example 3
[0090] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps:
[0091] The original magnetic field measurements are acquired in real time by a magnetometer, inertial navigation unit, and depth sensor mounted on the carrier. Carrier attitude angle angular velocity ,depth Based on the carrier location and the global geomagnetic model, the geomagnetic reference field was calculated. and its rate of change ;
[0092] Construct the covariance matrix of the interference model parameters Using the reinforcement learning agent as input, the system dynamically generates yaw angle increments while satisfying the maneuver constraints of the deep-sea operation platform. Pitch angle increment Incremental depth and stabilization time The sequence of maneuver commands;
[0093] Establish a carrier interference magnetic field model ,in, This is the hard magnetic interference vector. It is a soft magnetic induction matrix. For eddy current disturbance values, state-extended strong tracking Kalman filtering is used to jointly estimate model parameters. With background magnetic field And refine the parameters using least squares with L1 regularization;
[0094] Based on interference parameters Calculate the interfering magnetic field and from Subtract the mid-vector to output the corrected geomagnetic field. .
[0095] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A method for in-situ calibration and control of a deep-sea magnetometer, characterized in that, The method includes the following steps: S01. The original magnetic field measurement value is acquired in real time through the magnetometer, inertial navigation unit, and depth sensor mounted on the carrier. Carrier attitude angle angular velocity ,depth Based on the carrier location and the global geomagnetic model, the geomagnetic reference field was calculated. and its rate of change ; S02, Construct the covariance matrix of the interference model parameters. Using the reinforcement learning agent as input, the system dynamically generates yaw angle increments while satisfying the maneuver constraints of the deep-sea operation platform. Pitch angle increment Incremental depth and stabilization time The sequence of maneuver commands; S03. Establish a carrier interference magnetic field model. ,in, This is the hard magnetic interference vector. It is a soft magnetic induction matrix. For eddy current disturbance values, state-extended strong tracking Kalman filtering is used to jointly estimate model parameters. With background magnetic field And refine the parameters using least squares with L1 regularization; S04, Based on interference parameters Calculate the interfering magnetic field and from Subtract the mid-vector to output the corrected geomagnetic field. .
2. The in-situ calibration and control method for a deep-sea magnetometer according to claim 1, characterized in that, The reinforcement learning agent reward function in step S02 Designed as follows: ; in, The information entropy reward weight, Let be the parameter information entropy at time t. The parameter information entropy at time t-1, For the maximum permissible yaw angle, The maximum permissible pitch angle, For the maximum allowable depth, As the weight of maneuver cost, It is a negative exponential decay term for energy consumption. Energy consumption reward weight.
3. The in-situ calibration and control method for a deep-sea magnetometer according to claim 1, characterized in that, The step S03 of obtaining the regularized least squares refinement parameters includes the following steps: S31, State Extension Modeling: The background magnetic field... With interference parameters Jointly construct state vector The state equation is defined as ,in: This is the projection of the actual geomagnetic field onto the carrier coordinate system. Let be the angular velocity vector of the carrier at time k. Let k be the joint state vector at time k. Let k be the joint state vector when k-1. S32, Gradual Extinction Factor Adaptive Adjustment: ,in, For the innovation vector, To observe the Jacobian matrix, To predict the covariance matrix, R is the error value between the magnetometer and the pose sensor; S33, Covariance Reset: This enhances the ability to track parameter mutations, including: The process noise covariance matrix is... It is the variance matrix; S34. Constructing incremental equations: Where Z is the observed residual value, To increment the parameter The partial derivatives; S35, Solve the optimization problem with L1 constraint as follows: ,in, , The sum of the absolute values of the parameter increments.
4. The in-situ calibration and control method for a deep-sea magnetometer according to claim 1, characterized in that, The eddy current interference value The discretization implementation is as follows: ; in, The value of eddy current disturbance at time k-1. Let be the eddy current disturbance value at time k.
5. The in-situ calibration and control method for a deep-sea magnetometer according to claim 1, characterized in that, Steps S3 and S4 are implemented using a dual-thread architecture, specifically including: Background thread estimation: Execution cycle T est ≥1s, execute STEKF-RLS algorithm update ; Foreground compensation thread: Execution cycle T c ≤1ms, based on update Real-time computing and output ; Inter-thread data transfer: achieved through a circular buffer. The asynchronous update has a buffer depth of 1 to ensure that the compensation thread always gets the latest parameters without lock contention.
6. The in-situ calibration control method for a deep-sea magnetometer according to claim 1, characterized in that, The geomagnetic reference field in step S01 The calculation includes dynamic error correction: ; in, The static geomagnetic field calculated by the WMM global model. This is a real-time diurnal variation correction based on interpolation of data from the International Geomagnetic Network. This is the error feedback value.
7. The in-situ calibration and control method for a deep-sea magnetometer according to claim 1, characterized in that, The conditions for terminating the above-mentioned corrected geomagnetic field output in step S04 include: S41, Standard deviation of hard magnetic component ; Residual magnetic field fluctuations after S42's five consecutive maneuvers .
8. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the in-situ calibration control method for deep-sea magnetometers as described in any one of claims 1-7.
9. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the in-situ calibration control method for deep-sea magnetometers as described in any one of claims 1-7.
