A Multi-Source Fusion Relative Positioning Method and System for USV-ROVs Based on Cable Dynamic Constraints

By employing a positioning method that combines dynamic cable constraints with multi-source adaptive fusion, a dynamic parameter parabolic correction model and extended Kalman filtering are constructed. This solves the problems of deviation caused by cable bending and sagging and sensor noise sensitivity, achieving high-precision and stable underwater relative pose estimation and improving the robustness and navigation reliability of the system.

CN123065449APending Publication Date: 2026-11-10HANGZHOU DIANZI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611219234.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-12
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies for collaborative positioning between surface unmanned surface vessels and underwater remotely operated robots suffer from problems such as systematic deviations caused by mooring cable bending and sagging, sensitivity of geometric equations to sensor noise, and lack of adaptive adjustment capability of multi-source fusion filters, resulting in insufficient positioning accuracy and stability.

Method used

A positioning method combining cable dynamic constraints and multi-source adaptive fusion is adopted. By constructing a dynamic parameter parabolic correction model to eliminate cable sagging deviation, and by using extended Kalman filtering to introduce an adaptive covariance adjustment mechanism, high-precision underwater relative pose estimation is achieved.

Benefits of technology

It significantly improves underwater relative positioning accuracy and operational stability, reduces system cost and environmental sensitivity, and achieves highly robust navigation under conditions without underwater acoustic assistance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123065449A_ABST
    Figure CN123065449A_ABST
Patent Text Reader

Abstract

A collaborative localization method and system for USV-ROV based on cable dynamic constraints and multi-source adaptive fusion is proposed. The method includes: configuring a dual-antenna direction-finding real-time dynamic carrier phase differential module and a high-frequency inertial measurement unit at the end of the surface unmanned surface vessel; fusing the outputs of the real-time dynamic carrier phase differential module and the high-frequency inertial measurement unit to generate a global pose reference vector for the surface unmanned surface vessel; real-time acquisition of cable length, cable azimuth angle, and vertical depth difference; establishing the spatial position relationship between the underwater remotely operated vehicle (ROV) and the surface unmanned surface vessel by introducing dynamic equivalent parameters, and constructing a three-dimensional position observation vector of the ROV in the global reference coordinate system; based on the global pose reference vector of the surface unmanned surface vessel and the three-dimensional observation vector of the ROV in the global navigation coordinate system, introducing an adaptive observation noise covariance online adjustment strategy under the Kalman filter framework to generate the optimal state estimation vector and global position components of the ROV.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of marine robots, and in particular to a USV-ROV multi-source fusion relative positioning method and system based on cable dynamic constraints. Background Technology

[0002] Unmanned surface vehicles (USVs) are autonomous platforms capable of navigating on water surfaces and performing tasks such as inspection, monitoring, and obstacle avoidance. Remotely operated vehicles (ROVs) are underwater operational platforms within cross-domain collaborative systems. Equipped with an IMU and depth sensor, they connect to the USV via an umbilical cable to perform underwater exploration or data collection tasks. Cross-domain collaborative systems composed of USVs and ROVs are widely used in dam inspection, water quality monitoring, and three-dimensional exploration. Since satellite navigation signals cannot penetrate water, obtaining the relative pose between the surface and underwater platforms is a core prerequisite for achieving follow-along, autonomous recovery, and multi-vehicle collaborative operations. Underwater collaborative navigation is divided into two categories based on whether it relies on acoustic arrays: acoustic-assisted positioning and acoustic-free autonomous positioning. The former relies on underwater acoustic equipment such as ultra-short baselines or short baselines, while the latter attempts to utilize tethered cables (umbilical cables) connecting the surface and underwater for geometric ranging, thus eliminating rigid dependence on underwater acoustic hardware and becoming an important research direction for the rapid deployment of lightweight unmanned systems in the field.

[0003] One mainstream existing acoustic positioning scheme utilizes the physical information of the mooring line between the USV and ROV to construct a geometric ranging model. The basic idea is as follows: the USV obtains its absolute position via satellite navigation signals, and uses tension sensors, encoders, or angle sensors to measure the mooring line's extended length, pitch angle, and azimuth angle. Assuming the mooring line remains completely taut between the two ends (i.e., using an ideal straight baseline model), the horizontal span and depth difference between the ROV and the USV are directly calculated using trigonometric relationships, thus obtaining an estimate of the underwater platform's position. This method has low computational resource requirements, is easy to implement in real-time, and does not require additional underwater acoustic arrays, making it practical for static or low-flow-velocity operations.

[0004] However, this linear assumption scheme has three key drawbacks: First, the actual mooring line, due to its effective wet weight in water and the hydrodynamic resistance of the flow field, will inevitably exhibit a curved and drooping shape. Directly using the inverse linear solution will severely overestimate the horizontal span, introducing a system constant deviation of several meters. Second, during near-vertical diving or large curvature maneuvers, the geometric equations are extremely sensitive to sensor noise, and even small measurement fluctuations can cause severe oscillations in the calculated position. Third, the observation noise covariance of the subsequent multi-source fusion filter is usually fixed and lacks the ability to adaptively adjust to geometric distortions and innovation anomalies, resulting in trajectory jitter or even divergence under sudden disturbances, making it difficult to meet the requirements of high dynamics and high robustness in collaborative operations. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a USV-ROV cooperative positioning method and system based on cable dynamic constraints and multi-source adaptive fusion. This invention aims to completely eliminate expensive and interference-prone underwater acoustic positioning equipment, transforming the mooring line from a traditional physical source of interference into a "high-precision geometric observation source." It eliminates systematic deviations caused by cable sagging by constructing a dynamic parabolic correction model and utilizes a dual adaptive covariance online adjustment mechanism based on "geometric deviation" and "new information norm 2" to suppress numerical singularities and instantaneous sudden disturbances, achieving high-precision and robust underwater relative pose estimation under limited sensing conditions. This invention significantly improves the relative positioning accuracy, operational stability, and navigation control reliability of robot systems in complex aquatic environments.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect is a multi-source fusion relative positioning method for USV-ROVs based on cable dynamic constraints, which includes the following steps: S1. A dual-antenna direction-finding real-time dynamic carrier phase differential module and a high-frequency master inertial measurement unit are configured at the end of the unmanned surface vessel. The high-frequency master inertial measurement unit is used for one-step state prediction, and the low-frequency direction-finding real-time dynamic carrier phase differential module is used for state correction. The outputs of the two are fused to generate a continuous and drift-free global pose reference vector for the unmanned surface vessel.

[0007] S2. Through the follow-up mechanism and built-in follow-up IMU located below the cable outlet, relative attitude difference calculation is performed with the main inertial measurement unit of the surface unmanned vessel to obtain the cable outlet azimuth angle in real time to eliminate the interference of hull swaying; the cable length fed back by the winch encoder and the vertical depth difference calculated by the depth gauge of the underwater remotely operated robot are collected; the true equivalent horizontal span is solved online by constructing a parabolic geometric constraint equation with dynamic equivalent parameters, and combined with the cable outlet azimuth angle, vertical depth difference and global pose reference vector of the surface unmanned vessel, the three-dimensional position observation vector of the underwater remotely operated robot in the global reference coordinate system is constructed.

[0008] S3. Based on the global pose reference vector of the surface unmanned surface vessel and the three-dimensional observation vector of the underwater remotely operated vehicle in the global navigation coordinate system, a dual adaptive observation noise covariance online adjustment strategy including "geometric deviation channel weighting" and "new innovation L2 test" is introduced under the extended Kalman filter framework to generate the optimal state estimation vector and global position components of the underwater remotely operated vehicle.

[0009] Preferably, S1 includes: S11. Use a high-frequency inertial measurement unit to acquire high-frequency IMU data, including the planar position coordinates and heading angle of the unmanned surface vessel in the global reference coordinate system, and the longitudinal velocity, lateral velocity and yaw rate in the vehicle coordinate system; generate a state prediction vector based on the high-frequency IMU data and process noise vector through kinematic recursion. S12. Use the real-time dynamic carrier phase differential module to obtain low-frequency direction finding RTK data, including the planar position coordinates and heading angle of the underwater remotely operated robot at each moment; based on the low-frequency direction finding RTK data and the observation matrix of absolute position and heading angle extracted from the state prediction vector, generate a linear observation vector through transverse Mercator projection. S13. In the global reference coordinate system of the unmanned surface vessel, the extended Kalman filter is used to asynchronously fuse the state prediction vector and the linear observation vector to generate the global pose reference vector of the unmanned surface vessel.

[0010] Preferably, S2 includes: S21. Describe the location of the cable release point of the surface unmanned vessel and the location of the observation point of the underwater remotely operated robot in the global reference coordinate system, and obtain the cable length and vertical depth difference, and define the horizontal span. S22. Obtain the horizontal tension component of the cable and define characteristic parameters based on the effective wet weight per unit length. Based on the characteristic parameters, establish a coordinate system with the lowest point of the catenary as the local coordinate origin to obtain the standard catenary equation. Obtain the approximate parabola equation by performing a second-order Taylor expansion at the lowest point. S23. Construct a parabolic geometric constraint equation with dynamic equivalent parameters based on the parabolic approximation equation; obtain the true equivalent horizontal span by solving the parabolic geometric constraint equation online through numerical iteration method; and construct the three-dimensional position observation vector of the underwater remotely operated vehicle in the global reference coordinate system by combining the cable azimuth angle, the global pose reference vector of the surface unmanned vessel, and the vertical depth difference.

[0011] Preferably, S23 includes: The tangential tension of the cable at the water entry point is considered as a nominal quantity determined by the winch, and nominal geometric parameters are constructed. : In the formula, Set the output torque for the motor. Let the equivalent radius of the winch be 1. The overall transmission efficiency coefficient. Effective wet weight per unit length The nominal tangential tension of the cable at the exit point; considering the geometric projection relationship that the entry angle decreases as the diving depth of the underwater remotely operated vehicle increases, the geometric estimate of the entry angle is defined. satisfy To obtain dynamic equivalent parameters for: in, For vertical depth difference, To determine the cable length, the coordinate system is rewritten as a vertical plane coordinate system with the cable release point of the unmanned surface vessel as the origin and the depth direction downwards as positive, combined with endpoint constraints. and The equation of the parabolic trajectory is obtained as follows: in, The horizontal span.

[0012] As a preferred option, S23 also includes: Differentiating the equation of the parabolic trajectory yields the slope of the tangent line: Cable length With horizontal span Satisfying the implicit core geometric constraint equations: Within each sampling period, the implicit equations described above are solved online using the Newton-Raphson iteration method, and the true equivalent horizontal span is calculated in real time. .

[0013] As a preferred option, S23 also includes: azimuth angle of cable delivery The global heading angle of the unmanned surface vessel in the global pose reference vector of the unmanned surface vessel. Summing is performed to obtain the absolute azimuth angle of the cable. Combined with the current global coordinates of the unmanned surface vessel Constructing the three-dimensional observation vector of the underwater remotely operated robot in the global navigation coordinate system : .

[0014] Preferably, S3 includes: S31. Define the filtered state vector of the underwater remotely operated robot, and perform state prediction and state observation. S32. Design an adaptive online adjustment strategy for observation noise covariance that combines geometric bias and the new information L2 norm, and dynamically update the observation noise covariance matrix. S33. Calculate the Kalman gain matrix based on the adaptively corrected observation noise covariance matrix; complete the closed-loop update of the underwater remotely operated robot's filtered state vector and predicted covariance matrix using the Kalman gain matrix; continuously output the corrected optimal state estimation vector of the underwater remotely operated robot and extract its global position component.

[0015] Preferably, S31 includes: The filtered state vector includes the three-dimensional position and heading angle in the global navigation coordinate system, and the longitudinal velocity, lateral velocity, vertical translational velocity, and yaw rate in the body-fixed coordinate system. Combined with the yaw rate sampled by the high-frequency IMU at the underwater remotely operated robot end, the state is predicted using a discrete kinematics model. The first-order partial derivative of the state transition function with respect to the state estimate of the previous moment is obtained to derive the state transition Jacobian matrix, and the discrete recursion of the prediction covariance matrix is ​​completed. The planar position observations of the parabolic trajectory equation are extracted and combined with the vertical depth measurements provided by an independent depth sensor to construct the three-dimensional observation vector and linear observation equation of the underwater remotely operated robot.

[0016] Preferably, S32 includes: Geometric deviation index is constructed using the ideal assumption of perfect straightening and the equation of the parabolic trajectory. : in, For a moment The cable length measurement value obtained by the winch encoder. For a moment The vertical depth difference between the two ends is derived from the depth gauge of the underwater remotely operated vehicle and the surface reference datum of the unmanned surface vessel. For a moment The true equivalent horizontal span is calculated online using the implicit geometric constraint equation of the parabola with dynamic parameters; based on the geometric deviation. Online weighted adjustment of the variance of the planar position observation channel: In the formula, The variance of the nominal planar position observations, Weighted coefficients for geometric mapping; preserving depth channel variance Remaining unchanged, this forms the stage-specific observation noise covariance matrix: Calculate the planar position information vector: in, and For a moment The planar position observation components of the underwater remotely operated vehicle are derived from the geometric model of the parabolic cable with dynamic parameters. and For a moment One-step prediction component of the planar position of the underwater remotely operated vehicle calculated based on the discrete kinematics model; taking its second norm. As a discriminant; an adaptive decision threshold is set based on the current plane noise level. : In the formula, This is the threshold adjustment coefficient; during the update step, the following segmented control logic is executed: when When the observation is normal, the current observation noise covariance matrix is ​​maintained and a regular update is performed. when At that time, through the magnification factor Reduce the weight of anomalous observations in Kalman correction and generate an adaptively corrected observation noise covariance matrix. : A multi-source fusion relative positioning system for USV-ROVs based on cable dynamic constraints includes: unmanned surface vessels, cable management systems, and underwater remotely operated vehicles; A dual-antenna direction-finding real-time dynamic carrier phase differential module and a high-frequency inertial measurement unit are configured at the end of the unmanned surface vessel to realize state prediction and state correction, respectively. The outputs of the real-time dynamic carrier phase differential module and the high-frequency inertial measurement unit are fused to generate a continuous and drift-free global pose reference vector for the unmanned surface vessel. The cable management system includes a cable length measurement unit and a cable azimuth measurement unit, which are used to collect the cable length and cable azimuth, respectively; a depth gauge is used to collect the vertical depth difference relative to the water surface; by introducing dynamic equivalent parameters, the spatial position relationship between the underwater remotely operated robot and the surface unmanned vessel is established, and a three-dimensional position observation vector of the underwater remotely operated robot in the global reference coordinate system is constructed. Based on the global pose reference vector of the unmanned surface vessel and the three-dimensional observation vector of the underwater remotely operated robot in the global navigation coordinate system, an adaptive observation noise covariance online adjustment strategy is introduced under the Kalman filter framework to generate the optimal state estimation vector and global position components of the underwater remotely operated robot. The cable-based dynamic constraint-based USV-ROV multi-source fusion relative positioning system is used to implement the cable-based dynamic constraint-based USV-ROV multi-source fusion relative positioning method and its steps as described in the first aspect.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. Eliminates reliance on underwater acoustic arrays, significantly reducing costs and environmental sensitivity. Compared to traditional ROV acoustic positioning or surface array systems, this invention does not rely on expensive and cumbersome hydrophone arrays and underwater acoustic beacons. By transforming the mooring line from a traditional physical source of interference into a "high-confidence spatial geometric constraint source," it effectively solves the problems of sound wave multipath reflection, strong boundary interference, and exponential divergence of position calculated by pure IMU inertial navigation over time in confined waters. Under the premise of significantly reducing system hardware costs and deployment complexity, it achieves continuous, smooth, and drift-free estimation of underwater three-dimensional relative pose without underwater acoustic assistance.

[0018] 2. Introduce dynamic parabolic geometrical correction to completely eliminate systematic deviations in the suspension system. Compared with the traditional distance measurement model that simplifies the mooring line to an "ideal straight line", this invention constructs a dynamic parameter correction model based on a quasi-static parabola and moment-ingress angle projection. This model responds in real time to changes in cable tension and the angle of entry caused by diving depth, accurately reversing the true three-dimensional spatial bending morphology of the mooring line under the influence of flow field and its own weight. By numerically iteratively solving for the true horizontal span, it completely eliminates the constant meter-level position estimation bias caused by the straight-line model, significantly improving the geometric accuracy of the horizontal span and three-dimensional position calculations.

[0019] 3. It possesses a dual adaptive adjustment mechanism, which greatly improves filtering robustness and navigation stability. Compared to localization algorithms that employ fixed observation noise covariance (such as conventional EKF) or single weight switching, this invention proposes a dual adaptive online covariance adjustment strategy combining "geometric deviation channel weighting" and "new innovation L2 test". Under highly maneuverable or near-vertically sensitive conditions in ROVs, the algorithm can perceive the numerical uncertainty and sudden outlier disturbances of the geometric model in real time, dynamically adjusting the weight of planar position observations in Kalman correction. Without interrupting filter updates, it effectively suppresses the impact of geometric singularities and sudden measurement noise on state estimation, significantly improving the robot system's localization robustness and navigation stability in complex water environments.

[0020] 4. By integrating the entire technology chain of "sensing-correction-adaptation-closed loop", highly robust underwater acoustic relative positioning can be achieved. This invention constructs a complete technical chain: "Asynchronous global reference construction on the water surface—construction of relative pose vector based on attitude difference of servo IMU and cable length—physical correction calculation of dynamic parameter parabolic curves—dual adaptive covariance control based on geometric deviation decoupling and innovation L2 norm—underwater relative pose closed-loop output without underwater acoustic dependence." Each link is interconnected, not only eliminating orientation measurement disturbances caused by hull rolling at the hardware level, but also fundamentally solving the problems of systematic deviations and geometrically singular numerical oscillations in the suspended system from the perspective of physical modeling and adaptive estimation. Under limited sensing conditions, it achieves high-precision, continuous, and drift-free underwater relative pose closed-loop estimation without underwater acoustic dependence, significantly improving the operational stability and navigation reliability of heterogeneous unmanned systems in complex and confined waters. Attached Figure Description

[0021] Figure 1 This is a diagram illustrating the overall framework of USV-ROV cooperative positioning in Embodiment 1 of the present invention. Figure 2 This is a schematic diagram of the asynchronous update mechanism of the USV water surface reference datum in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the relative positioning of a USV-ROV based on cable measurement information in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the AEKF fusion positioning framework and the online adjustment mechanism of observation noise covariance in Embodiment 1 of the present invention; Figure 5 This is a model diagram of the USV-ROV cross-domain collaborative system in Embodiment 2 of the present invention; Figure 6 This is a structural model of the cable management system according to Embodiment 2 of the present invention; Figure 7 This is a cross-sectional view of the cable length measuring unit and cable take-up structure in Embodiment 2 of the present invention; Figure 8 This is a schematic diagram of the cable follower mechanism in Embodiment 2 of the present invention. Detailed Implementation

[0022] To make the technical means, inventive features, objectives, and effects of this invention readily understandable, the invention is further described below in conjunction with specific embodiments. However, this invention is not limited to the following embodiments.

[0023] It should be noted that the process sequence, module ratio, number of nodes, and parameter values ​​shown in the diagram are only for illustrating the implementation of the present invention and are not intended to limit the scope of the present invention. Any equivalent substitutions made to the model type, encoder structure, language model type, retrieval model, database format, field name, threshold size, or report template without departing from the core concept of the present invention should fall within the scope of the technical solutions disclosed in the present invention.

[0024] This invention does not simply protect a certain existing visual model, language model, or database retrieval model. Instead, it expands the diagnosis of tomato leaf diseases from a single image classification to an intelligent decision-making process with evidence tracing, differential diagnosis, and security management capabilities by constructing a unified case representation, mechanism profiling and causal chain, positive and negative example evidence diversion retrieval, shared blackboard multi-intelligent diagnosis, and security review mechanism.

[0025] The technical problem to be solved by this invention is: addressing the core issues of acoustic positioning failure in shallow and confined waters, large deviations in the traditional straight baseline assumption, and numerical sensitivity of geometric inverse solutions under extreme conditions, the invention focuses on mooring dynamic constraint modeling, multi-source state estimation, and adaptive calculation and optimization of filter variance to achieve high-precision acoustic relative positioning of underwater remotely operated vehicles facing unmanned surface vessels.

[0026] Example 1: like Figure 1 As shown, the USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints specifically includes the following three progressive technical steps: Step S1: Asynchronous construction of global water surface reference benchmark To generate the global reference benchmark required for underwater relative positioning and cooperative navigation, the UAV is equipped with a dual-antenna direction-finding real-time dynamic carrier phase differential system (RTK) and a high-frequency inertial measurement unit (IMU) as the surface reference benchmark sensing hardware. The RTK, combined with the dual-antenna direction-finding module, is used to acquire the USV's high-precision absolute geographical location and a reference heading angle without cumulative drift. The IMU, containing accelerometers and gyroscopes, is used to acquire the angular velocity and acceleration information of both the USV and ROV at high frequency, providing motion state prediction and attitude recursion. The output of the direction-finding RTK is projected using Universal Transverse Mercator (UTM). Latitude and longitude location information interest The coordinates are converted to two-dimensional planar coordinates, and a global reference coordinate system (North-East-Down Frame, NED) is established within the operating area. The origin of this global reference coordinate system is set at the operating water area, and the three positive axes point to due north, due east, and vertically downwards, respectively. The axis points due north. The axis points due east, and the USV's heading angle is... The positive direction of the axis is the reference datum, and the counterclockwise direction is positive, denoted as . .

[0027] In the global reference coordinate system of the unmanned surface vessel, the extended Kalman filter (EKF) is used to asynchronously fuse RTK and IMU data. The specific construction process is as follows: S1.1 Definition of State Vector and State Prediction Considering the planar motion characteristics of the USV on the water surface, define the discrete time intervals. Filtered state vector of unmanned surface vessel for: In the formula, This indicates the planar position coordinates and heading angle of the USV in the global reference coordinate system; This represents the longitudinal velocity, lateral velocity, and yaw rate of the USV in the vehicle coordinate system. The filtered output USV global pose reference vector is defined as follows: Discrete time Based on the state estimation, combined with the high-frequency yaw rate measurements acquired by the IMU. Based on kinematic recursion, a one-step state prediction is performed, and its discrete prediction equation is expressed as: In the formula, The process noise vector represents the noise from wind and waves, flow field effects, and unmodeled errors, and follows a Gaussian distribution. ,in Let be the process noise covariance matrix.

[0028] S1.2 Direction Finding RTK Observation and State Correction First, using the Universal Transverse Mercator (UTM) projection algorithm, the direction-finding RTK is projected at time [time value missing]. The received WGS-84 latitude and longitude absolute geographic coordinates are converted into plane rectangular coordinates in the Global Reference Coordinate System (NED). .

[0029] The heading angle output by the dual-antenna direction finding module , building time Direction finding RTK plane measurement vector (subscript) (Represents the sensor's raw measurement value): The linear observation equations for direction-finding RTK are as follows: In the formula, The observed noise vector follows a Gaussian distribution. , The observation noise covariance matrix is ​​determined by the accuracy of the RTK plane position and the accuracy of the heading measurement. For use from the state vector The observation matrix for extracting the absolute position coordinates and heading angle of the plane is expressed as: S1.3 Asynchronous Update Mechanism and Reference Output The overall update process is as follows: Figure 2 As shown. In practical systems, the data update frequency of the IMU is usually higher than that of the direction-finding RTK. To balance short-term continuity and position correction capability, an asynchronous update method is used to process two types of data: state prediction is performed when IMU data arrives, and observation correction is performed when direction-finding RTK data arrives. This step uses an asynchronous update mechanism for state estimation: Prediction step: When high-frequency IMU data arrives, the yaw rate measurement is used. The execution state is predicted in one step to maintain the high-frequency continuity of the USV reference pose output within the interval between two RTK update; Correction step: When low-frequency direction-finding RTK data arrives, the observation vector is called. With observation matrix Perform Kalman gain calculation and state correction to correct the cumulative motion drift of the one-step prediction and suppress long-term position error divergence.

[0030] Through an asynchronous update mechanism, the USV's global pose reference vector is continuously output, providing a smooth and drift-free output. This serves as a unified reference input for subsequent dynamic geometric constraint calculation of cables and adaptive fusion positioning of underwater ROVs.

[0031] Step S2: Dynamic parameter parabolic cable-constrained observation model A schematic diagram of the relative positioning of a USV-ROV based on cable measurement information is shown below. Figure 3 As shown, this step uses the cable length measurement value collected by the cable management system. azimuth of cable output And the measured vertical depth difference between the two ends calculated from the ROV depth gauge and the water surface reference datum. This establishes the spatial relationship between the ROV and the USV to supplement underwater positioning information. The specific construction process is as follows: S2.1 Geometric Relationships and Linear Model First, describe the location of the USV departure point in the global reference coordinate system (NED). Location of ROV observation point The relative positional relationship between them. The relative displacement between the two is defined as: Wherein, the vertical depth difference is defined as Horizontal span is defined as .

[0032] If we disregard the sag of the cable due to its own weight, buoyancy, and fluid forces, and approximate the cable as an ideal straight line connecting two points, we can obtain an estimate of the horizontal span under the straight baseline model based on the Pythagorean theorem. for: In the formula, This refers to the cable length fed back by the winch encoder. In actual underwater operations, the cable will experience sag bending due to its own effective wet weight, and the actual horizontal span must meet the following requirements. If a linear model is used directly... Performing geometric conversions will overestimate the horizontal span of the ROV and introduce meter-level systematic biases, so a parabolic correction model needs to be introduced.

[0033] S2.2 Quasi-static Parabolic Correction Model Under quasi-static and actively tensioned conditions, the physical shape of a flexible cable in the vertical plane can be described by a catenary. Let the horizontal tension component of the cable be... The effective wet weight per unit length is Define feature parameters Establish a coordinate system with the lowest point of the catenary as the local origin. The standard equation of the catenary is expressed as: Under conditions where the cable remains taut and has minimal sag, it is possible to to Performing a local Taylor expansion and retaining only the second-order terms, we obtain the approximate equation of a local second-order parabola: The comparison shows that the second-order parabola approximation retains the main drooping characteristics of the catenary and significantly reduces the geometric conversion error compared to the straight-line model, but with fixed parameters... It cannot reflect the dynamic changes in tension and sag under different operating conditions, and further dynamic parameter correction is needed.

[0034] S2.3 Dynamic Parameter Correction Model Based on the winch torque control characteristics, the tangential tension of the cable at the water entry point is approximated as a nominal quantity determined by the winch, and nominal geometric parameters are introduced. : In the formula, Set the output torque for the motor. Let the equivalent radius of the winch be 1. To determine the overall transmission efficiency coefficient, considering the geometric projection relationship of the decreasing entry angle as the ROV's diving depth increases, a geometric estimate of the entry angle is defined. satisfy To obtain dynamic equivalent parameters for: The coordinate system is rewritten as a vertical plane coordinate system with the USV departure point as the origin and the depth direction downwards as positive, combined with endpoint constraints. and The equation of the parabolic trajectory is obtained as follows: right Differentiate to get the slope of the tangent line Based on the arc length infinitesimal element relationship, the measured value of the cable laying length. Horizontal span to be determined Satisfying the implicit core geometric constraint equations: Within each sampling period, the implicit equations described above are solved online using the Newton-Raphson iteration method, and the true equivalent horizontal span is calculated in real time. .

[0035] The cable azimuth angle calculated by the servo measurement unit and the global heading angle of the unmanned surface vessel output in step S1 Calculate the absolute azimuth angle of the cable in the global coordinate system. .

[0036] Finally, combining the current global coordinates of the unmanned surface vessel... Constructing the three-dimensional observation vector of the underwater remotely operated robot in the global navigation coordinate system : In the formula, for Figure 3 The azimuth angle of the output cable calculated by the attitude difference decomposition between the servo IMU and the USV master IMU. For the global heading angle of the unmanned surface vessel, The true equivalent horizontal span for solving the parabolic model. This represents the vertical depth difference.

[0037] 3D observation vector of unmanned surface vessel The physical correction of the cable sag error was completed, serving as the baseline input for subsequent adaptive extended Kalman filter (AEKF) state updates.

[0038] Step S3: Fusion localization based on adaptive extended Kalman filter (AEKF) The global reference pose of the unmanned surface vessel constructed in step S1 Combined with the dynamic cable geometry plane observations derived in step S2, and data from the high-frequency IMU and depth sensor on the ROV, an adaptive online adjustment strategy for observation noise covariance, incorporating "geometric bias adjustment" and "new information norm 2 test," is introduced within the Extended Kalman Filter (EKF) framework. Figure 4 As shown, a smooth and high-precision estimation of the three-dimensional pose and motion state of an ROV underwater is achieved without the assistance of an underwater acoustic array. The specific construction process is as follows: S3.1 Prediction and Observation Model Construction Considering the three-dimensional translational and yaw rotational characteristics of an ROV underwater, we define discrete time intervals. ROV filter state vector for: In the formula, This indicates the ROV's three-dimensional position in the global navigation coordinate system (NED). This refers to the ROV's heading angle. This indicates the longitudinal, lateral, and vertical translational velocities of the ROV in the body-fixed coordinate system. This indicates the yaw rate.

[0039] (1) State prediction stage Set the filter sampling period to Combined with the yaw rate sampled by the high-frequency IMU at the ROV end. One-step state propagation is performed using a discrete kinematics model: In the formula, The process noise vector follows a Gaussian distribution. , This is the process noise covariance matrix, used to characterize the uncertainties of model simplification errors, water flow field disturbances, and thruster output fluctuations. Its specific position and heading state propagation equation is expressed as: Assuming a short sampling period The internal velocity and yaw rate remain relatively flat and constant, that is... For the state transition function Taking the first-order partial derivative of the state estimate from the previous time step, we can derive the state transition Jacobian matrix. Complete the discrete recursion of the predicted covariance matrix: (2) State observation phase Extract the planar position observation components derived from the dynamic parameter parabolic cable model in step S2. Combined with vertical depth measurements provided by an independent pressure-type depth sensor. Building ROV at all times Three-dimensional measurement vector for: The linear observation equation is constructed as follows: In the formula, Let the measurement noise vector have a mean of zero and a covariance matrix of... The Gaussian distribution, i.e. ,in This refers to the adaptive observation noise covariance matrix that is dynamically adjusted online in step S3.2; For use from ROV state vector The observation matrix from which the three-dimensional position components are extracted is expressed as: S3.2 Online Adjustment Strategy for Observation Noise Covariance To address the issue of decreased reliability in geometric observations caused by cable slack, morphological bending, and sudden underwater disturbances, this step designs an online adjustment mechanism for observation weights that combines geometric bias and the innovation L2 norm, dynamically updating the observation noise covariance matrix. : (1) Geometric deviation channel weighted adjustment A geometric deviation index is constructed using the ideal assumption of complete straightening and the values ​​calculated by the parabolic geometric model in step S2. : In the formula, For a moment The cable length measurement value obtained by the winch encoder For a moment The vertical depth difference measurement between the two ends is derived from the depth gauge of the underwater remotely operated vehicle and the water surface reference datum. For a moment The true equivalent horizontal span is calculated based on the solution of the implicit geometric constraint equation of the parabolic curve with dynamic parameters.

[0040] Based on this geometric deviation Online weighted adjustment of the variance of the planar position observation channel: In the formula, and Representing time respectively Dynamically adjusted shaft and Variance of observed position in the axial plane; The variance of the nominal planar position observations, These are the weighting coefficients for the geometric mapping. Since depth information is provided by independent high-precision depth gauges, the depth channel variance is preserved. Remaining unchanged, this forms the stage-specific observation noise covariance matrix: (2) Adaptive weighted adjustment of the new innovation L2 norm To suppress the impact of transient underwater disturbances on the filtered estimation, the planar position innovation vector is calculated: In the formula and Representing time respectively The results calculated from the dynamic parabolic cable geometry model shaft and Axial plane position observation components; and Representing time respectively Based on the discrete kinematics model, the following is derived shaft and One-step prediction of axial plane position components.

[0041] And take its second norm. As a discriminant, an adaptive decision threshold is set based on the current plane noise level. : In the formula, This is the threshold adjustment coefficient. During the update step, the following segmented control logic is executed: when When the time is right, it indicates that the observation is normal and the current state should be maintained. Perform a routine update; when When this occurs, it indicates the presence of a sudden, abnormal disturbance. Instead of directly removing observations to prevent update interruption, the amplification factor is used. Reduce the weight of this anomalous observation in the Kalman correction to generate an adaptively corrected observation noise covariance matrix. : In the formula: The adaptive amplification factor represents the observation noise covariance; This represents the depth channel observation variance that remains constant across independent depth sensors; This represents the observation noise covariance matrix used to calculate the Kalman gain after adaptive correction.

[0042] S3.3 Kalman Gain Solution and State-Closed-Loop Correction The adaptively corrected observation noise covariance matrix generated in step S3.2 Substituting into the Kalman gain matrix solution formula, calculate the time step. Kalman gain matrix for: In the formula: Indicates time of Vikalman gain matrix; Indicates time The state prediction covariance matrix; The observation matrix is ​​extracted to represent the three-dimensional position state. This represents the 3×3 dimensional observation noise covariance matrix after adaptive correction by combining geometric bias and the new information norm in step S3.2.

[0043] This completes the closed-loop update of the filtered state vector and predicted covariance matrix of the remotely operated vehicle (ROV): In the formula: Indicates time One-step prediction vector of ROV state derived from kinematic model; This represents a three-dimensional observation vector constructed from a dynamic parabolic geometric model and a depth gauge. Represents a measurement information vector containing the three-dimensional position; Indicates time The optimal state estimation vector of the ROV after closed-loop correction; Indicates time Corrected state estimation covariance matrix; express 3D identity matrix.

[0044] The system continuously outputs the corrected optimal state estimation vector of the underwater remotely operated robot. Extract the global position component. It achieves highly robust, continuous, and drift-free relative positioning under complex water flow and dynamic changes in overhang morphology.

[0045] Example 2: like Figure 5The USV-ROV multi-source fusion relative positioning system shown consists of three parts: a surface unmanned vessel, a cable management system, and an underwater remotely operated robot.

[0046] As the main control and reference anchor point on the water surface, the unmanned surface vessel is equipped with a real-time dynamic carrier phase differential system and a high-frequency inertial measurement unit. It uses a dual-antenna direction finding module to eliminate heading drift and provide a robust global absolute reference.

[0047] The underwater remotely operated vehicle is equipped with a high-performance 3D IMU (to record the vehicle's motion trajectory), a high-precision pressure depth gauge (to obtain the relative depth difference H to the water surface), and a power thruster. It establishes a high-bandwidth communication link with the USV via an umbilical cable.

[0048] like Figure 5 As shown, the cable management system is a crucial connection unit between the USV and ROV, primarily responsible for two functions: cable deployment and retrieval, and cable status measurement. This invention integrates the deployment mechanism and cable length measurement unit into a single unit with a compact layout, installed near the geometric center of the USV deck to minimize the impact of cable tension changes on the platform's attitude during deployment and retrieval.

[0049] The cable management system needs to simultaneously acquire cable length and cable departure direction to characterize the cable geometry between the USV and ROV. Based on this, this invention designs a two-part structure: a cable length measurement unit and a cable departure azimuth angle measurement unit. For example... Figure 6 , Figure 7 As shown, the cable length measurement unit is located at the cable exit end of the winch and mainly consists of a guide wheel and a high-resolution encoder. After the cable is drawn from the winch, it adheres to the guide wheel and passes downwards under the guidance of the structure. During the winding and unwinding process, the cable moves synchronously with the guide wheel. The system converts pulse counting into the guide wheel rotation angle and combines this with the guide wheel circumference to obtain the cable length. L The cumulative value. Since the measurement position is set at the cable outlet, the diameter change caused by the overlapping of the winch cable will not be included in the length calculation. Therefore, this method has better measurement stability and is more suitable for long-term continuous operation.

[0050] like Figure 7 , Figure 8 As shown, to obtain the cable departure azimuth angle, the cable management system has a follow-up mechanism installed below the cable departure port. This mechanism can swing freely in the horizontal direction. When the ROV's movement causes a change in the direction of force on the cable, the follow-up mechanism will deflect synchronously with the cable direction. An angle sensor IMU is installed inside the follow-up mechanism to measure its horizontal deflection angle relative to the USV's hull coordinate system, and this angle is defined as the cable departure azimuth angle. (in the ship's coordinate system) x (Using the axis as a reference, counterclockwise is positive).

[0051] Therefore, the cable length The azimuth angle of the cable output calculated in real time by the servo IMU Together, they reflect the relative spatial geometry between the USV and ROV. Among them, cable length... Used to describe the distance change between the two platforms, cable azimuth. These quantities are used to describe the planar orientation of the cable relative to the USV hull. In subsequent cooperative positioning, these two quantities will serve as important measurement inputs for the dynamic parabolic geometry model and adaptive extended Kalman filter; in cable retrieval and deployment control, they can also be used to characterize the cable's stress state and assist in retrieval and deployment control adjustments.

[0052] The system also includes: a surface reference calculation unit (deployed on the USV end to acquire dual-antenna RTK and high-frequency main IMU data), a cable management integration unit (including a winch deployment and take-up mechanism with encoder, a suspension follow-up mechanism below the cable outlet, and a built-in follow-up IMU, used to specifically measure the relative position and distance of the ROV to the USV), a cable geometry solving unit (deployed on the main control end to collect cable length, follow-up IMU attitude and depth difference and solve the parabolic equation), an underwater adaptive fusion positioning unit (deployed on the ROV end main control board), and a distributed communication interaction unit (based on the ROS communication framework).

[0053] System software design includes: The software system adopts a distributed node deployment, relying on the Robot Operating System (ROS) to achieve cross-node data interaction and collaboration; communication and timing nodes uniformly align the time scales of each sensor to ensure the timing synchronization of RTK, IMU, and cable sensor data; the observation state calculation node is responsible for processing the input raw physical quantities ( The algorithm maps the calculated 3D pose information to the defined geometrically constrained observation model; the filtering and control nodes, as the core algorithm terminals, execute the adaptive extended Kalman filter (AEKF) and process the calculated 3D pose information. The commands are converted into thruster drive commands, and the relative positioning and navigation control are completed in a closed loop.

[0054] The inventiveness of this invention lies in: 1. Cross-media global coordinate reference and high- and low-frequency asynchronous fusion construction mechanism.

[0055] This invention deploys a dual-antenna direction-finding RTK module and a high-frequency IMU at the end of an unmanned surface vessel (ROV), establishing a global reference coordinate system based on UTM coordinate transformation. The invention employs extended Kalman filtering to asynchronously fuse the high-frequency IMU and low-frequency direction-finding RTK data: upon arrival of the high-frequency IMU, a one-step state prediction is performed to maintain the continuity of pose output; upon arrival of the low-frequency RTK, the measurement vector is invoked to perform state correction to eliminate accumulated inertial navigation drift. The protection point emphasizes a construction method of "establishing a unified reference across media, asynchronous recursion and correction of high and low frequency sensors, and continuous drift-free surface reference pose output," providing a high-precision and smooth spatiotemporal reference for the underwater ROV.

[0056] 2. A geometrical correction mechanism for relative position based on attitude difference of servo IMU and dynamic parabola.

[0057] This invention transforms the mooring line connecting the USV and ROV from a traditional physical interference source into a "high-reliability spatial geometric constraint source." At the hardware measurement level, this invention utilizes the heading angle of a servo IMU within a servo mechanism to obtain the azimuth angle of the outgoing cable. Combined with the length of the cable This is specifically designed to construct the relative spatial geometric constraints between the ROV and the USV. At the physical geometry solution level, a dynamic parameter correction model based on torque and entry angle projection is constructed. Using the following implicit core geometric constraint integral equation: The system uses a numerical iterative method to calculate the true relative horizontal span online. Furthermore, this is combined with the cable azimuth angle. With vertical depth difference The three-dimensional relative position vector of ROV to USV is calculated, and the global three-dimensional observation vector of ROV is constructed by superimposing the absolute position of USV.

[0058] 3. A dual adaptive covariance adjustment mechanism based on geometric bias and novel second norm.

[0059] This invention constructs a dual adaptive observation noise covariance online adjustment strategy that incorporates both "geometric deviation" and "novel L2 norm" during the underwater ROV fusion update stage. First, the geometric deviation is constructed using the ideal perfect straightening assumption and the solution value of the parabolic model. This serves as a mapping index for the uncertainty of planar geometric observations, and only the noise variance of the planar position observation channel is dynamically and incrementally weighted. While keeping the variance of the independent depth sensor channels unchanged, the planar position innovation vector is extracted and its L2 norm is calculated. A dynamic adaptive decision threshold is established based on the current plane noise level. When the innovation norm 2 exceeds the threshold, an adaptive amplification factor is used to dynamically amplify the observation noise covariance, achieving flexible weighting of sudden outliers and instantaneous underwater disturbances instead of direct removal, ensuring continuous and uninterrupted filtering updates. The core of this mechanism lies in the fact that geometric bias decoupling control only applies to planar channels affected by cables, while the innovation norm 2 defends against sudden outliers, achieving adaptive weighting and uninterrupted updates without changing the observation model itself.

[0060] 4. An AEKF state closed-loop estimation and dynamic update mechanism for cross-medium underwater collaboration.

[0061] This invention, during operations not yet reaching the target point or during continuous collaborative work, performs real-time IMU high-frequency state prediction, cable geometry and depth observation construction, dual adaptive covariance matching, and state closed-loop correction at the ROV end master control. In the prediction phase, 3D kinematic state propagation and covariance recursion are performed using ROV-end IMU measurement data. In the update phase, the Kalman gain is calculated using the observation noise covariance matrix after the aforementioned dual adaptive adjustment, and closed-loop correction is performed on the ROV's 3D position and 3D translational velocity, continuously outputting a smooth, drift-free underwater ROV global position estimate. When there is a sudden change in cable length, USV surface reference re-initialization, or excessive uncertainty in geometric calculations under near-vertical conditions is detected, the system can automatically adjust the fusion weights or trigger a local state reset, ensuring that the underwater relative positioning algorithm always obtains a highly robust estimate consistent with the current system motion state and environmental resistance characteristics. This protects the algorithm's closed-loop state update process, adaptive adjustment logic, and dynamic update relationship between surface and underwater collaborative positioning.

[0062] 5. Systematic implementation and program carrier protection for the USV-ROV collaborative platform.

[0063] This invention also protects the USV-ROV cooperative positioning system and its programmed implementation form that implements this relative positioning method. This system is completely independent of expensive and easily interfered-with underwater acoustic positioning equipment (such as USBL / SBL / LBL), and its hardware and software architecture may include: a surface reference calculation unit (deployed on the USV side to acquire dual-antenna RTK and high-frequency main IMU data), a cable management integration unit (including a winch deployment and take-up mechanism with an encoder, a suspended follow-up mechanism below the cable outlet, and a built-in follow-up IMU, used specifically to measure the relative orientation and distance of the ROV relative to the USV), a cable geometry solving unit (deployed on the main control end to collect cable length, follow-up IMU attitude and depth difference, and solve the parabolic equation), an underwater adaptive fusion positioning unit (deployed on the ROV side main control board), and a distributed communication interaction unit (based on the ROS communication framework). ROS stands for Robot Operating System. The distributed middleware framework used in the software system of this invention is used to realize the calculation of high-frequency data across nodes, clock synchronization, and command interaction. This protection point is used to cover the system and program implementation forms of this method in shipborne computing platforms, underwater robot embedded main control systems, shore-based monitoring systems, and underwater collaborative simulation platforms.

[0064] Of the above-mentioned inventive points, (1) constitutes the spatiotemporal reference layer for cross-media collaboration of the present invention; (2) and (3) constitute the core protection layer of the present invention that distinguishes it from conventional "ideal straight-line ranging" or "fixed covariance EKF"; (4) constitutes the underwater state closed-loop estimation and anti-disturbance update protection layer; and (5) is used to cover the system hardware architecture and programmatic implementation. The protection focus of the present invention is not only "combining cable ranging with EKF" or "adaptively adjusting the covariance of Kalman filtering", but also "constructing an asynchronous global reference on the water surface - constructing a relative pose vector based on the attitude difference of the follow-up IMU and the cable length - solving the physical correction of the dynamic parameter parabola - dual adaptive covariance control based on geometric deviation decoupling and the novel second norm - underwater relative pose closed-loop output without underwater acoustic dependence".

Claims

1. A multi-source fusion relative positioning method for USV-ROV based on cable dynamic constraints, characterized in that, Includes the following steps: S1. A dual-antenna direction-finding real-time dynamic carrier phase differential module and a high-frequency inertial measurement unit are configured at the end of the unmanned surface vessel to realize state prediction and state correction, respectively. The outputs of the real-time dynamic carrier phase differential module and the high-frequency inertial measurement unit are fused to generate a continuous and drift-free global pose reference vector for the unmanned surface vessel. S2. Real-time acquisition of cable length, cable azimuth angle, and vertical depth difference calculated by ROV depth gauge and water surface reference. By introducing dynamic equivalent parameters, the spatial positional relationship between the underwater remotely operated robot and the surface unmanned vessel is established, and the three-dimensional position observation vector of the underwater remotely operated robot in the global reference coordinate system is constructed. S3. Based on the global pose reference vector of the surface unmanned surface vessel and the three-dimensional observation vector of the underwater remotely operated robot in the global navigation coordinate system, an adaptive observation noise covariance online adjustment strategy is introduced under the Kalman filter framework to generate the optimal state estimation vector and global position components of the underwater remotely operated robot.

2. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 1, characterized in that, S1 includes: S11. Use a high-frequency inertial measurement unit to acquire high-frequency IMU data, including the planar position coordinates and heading angle of the unmanned surface vessel in the global reference coordinate system, and the longitudinal velocity, lateral velocity and yaw rate in the vehicle coordinate system. Based on high-frequency IMU data and process noise vector, a state prediction vector is generated through kinematic recursion. S12. Use the real-time dynamic carrier phase differential module to obtain low-frequency direction finding RTK data, including the planar position coordinates and heading angle of the underwater remotely operated robot at each moment; Based on low-frequency direction-finding RTK data and the observation matrix that extracts absolute position and heading angle from the state prediction vector, a linear observation vector is generated by transverse Mercator projection. S13. In the global reference coordinate system of the unmanned surface vessel, the extended Kalman filter is used to asynchronously fuse the state prediction vector and the linear observation vector to generate the global pose reference vector of the unmanned surface vessel.

3. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 1, characterized in that, S2 include: S21. Describe the location of the cable release point of the surface unmanned vessel and the location of the observation point of the underwater remotely operated robot in the global reference coordinate system, and obtain the cable length and vertical depth difference, and define the horizontal span. S22. Obtain the horizontal tension component of the cable and define characteristic parameters based on the effective wet weight per unit length. Based on the characteristic parameters, establish a coordinate system with the lowest point of the catenary as the local coordinate origin to obtain the standard catenary equation. Obtain the approximate parabola equation by performing a second-order Taylor expansion at the lowest point. S23. Construct a parabolic geometric constraint equation with dynamic equivalent parameters based on the parabolic approximation equation; obtain the true equivalent horizontal span by solving the parabolic geometric constraint equation online through numerical iteration method; and construct the three-dimensional position observation vector of the underwater remotely operated vehicle in the global reference coordinate system by combining the cable azimuth angle, the global pose reference vector of the surface unmanned vessel, and the vertical depth difference.

4. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 3, characterized in that, S23 includes: The tangential tension of the cable at the water entry point is considered as a nominal quantity determined by the winch, and nominal geometric parameters are constructed. : In the formula, Set the output torque for the motor. Let the equivalent radius of the winch be 1. The overall transmission efficiency coefficient. Effective wet weight per unit length The nominal tangential tension of the cable at the exit point; considering the geometric projection relationship that the entry angle decreases as the diving depth of the underwater remotely operated vehicle increases, the geometric estimate of the entry angle is defined. satisfy To obtain dynamic equivalent parameters for: in, For vertical depth difference, To determine the cable length, the coordinate system is rewritten as a vertical plane coordinate system with the cable release point of the unmanned surface vessel as the origin and the depth direction downwards as positive, combined with endpoint constraints. and The equation of the parabolic trajectory is obtained as follows: in, The horizontal span.

5. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 4, characterized in that, S23 also includes: Differentiating the equation of the parabolic trajectory yields the slope of the tangent line: Cable length With horizontal span Satisfying the implicit core geometric constraint equations: Within each sampling period, the implicit equations described above are solved online using the Newton-Raphson iteration method, and the true equivalent horizontal span is calculated in real time. .

6. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 5, characterized in that, S23 also includes: azimuth angle of cable delivery The global heading angle of the unmanned surface vessel in the global pose reference vector of the unmanned surface vessel. Summing is performed to obtain the absolute azimuth angle of the cable. Combined with the current global coordinates of the unmanned surface vessel Construct the three-dimensional observation vector of the underwater remotely operated robot in the global navigation coordinate system. : 。 7. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 1, characterized in that, S3 include: S31. Define the filtered state vector of the underwater remotely operated robot, and perform state prediction and state observation. S32. Design an adaptive online adjustment strategy for observation noise covariance that combines geometric bias and the new information L2 norm, and dynamically update the observation noise covariance matrix. S33. Calculate the Kalman gain matrix based on the adaptively corrected observation noise covariance matrix; complete the closed-loop update of the underwater remotely operated robot's filtered state vector and predicted covariance matrix using the Kalman gain matrix; continuously output the corrected optimal state estimation vector of the underwater remotely operated robot and extract its global position component.

8. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 7, characterized in that, S31 includes: The filtered state vector includes the three-dimensional position and heading angle in the global navigation coordinate system, and the longitudinal velocity, lateral velocity, vertical translational velocity, and yaw rate in the body-fixed coordinate system. Combined with the yaw rate sampled by the high-frequency IMU at the underwater remotely operated robot end, the state is predicted using a discrete kinematics model. The first-order partial derivative of the state transition function with respect to the state estimate of the previous moment is obtained to derive the state transition Jacobian matrix, and the discrete recursion of the prediction covariance matrix is ​​completed. The planar position observations of the parabolic trajectory equation are extracted and combined with the vertical depth measurements provided by an independent depth sensor to construct the three-dimensional observation vector and linear observation equation of the underwater remotely operated robot.

9. The USV-ROV multi-source fusion relative positioning method based on cable dynamic constraints according to claim 7, characterized in that, S32 includes: Geometric deviation index is constructed using the ideal assumption of perfect straightening and the equation of the parabolic trajectory. : in, For a moment The cable length measurement value obtained by the winch encoder For a moment The vertical depth difference between the two ends is derived from the depth gauge of the underwater remotely operated vehicle and the surface reference datum of the unmanned surface vessel. For a moment The true equivalent horizontal span is calculated online numerically based on the implicit geometric constraint equation of the parabola with dynamic parameters; based on geometric deviation. Online weighted adjustment of the variance of the planar position observation channel: In the formula, The variance of the nominal planar position observations, Weighted coefficients for geometric mapping; preserving depth channel variance Remaining unchanged, this forms the stage-specific observation noise covariance matrix: Calculate the planar position information vector: in, and For a moment The planar position observation components of the underwater remotely operated vehicle are derived from the geometric model of the parabolic cable with dynamic parameters. and For a moment One-step prediction component of the planar position of the underwater remotely operated vehicle calculated based on the discrete kinematics model; taking its second norm. As a discriminant; an adaptive decision threshold is set based on the current plane noise level. : In the formula, This is the threshold adjustment coefficient; during the update step, the following segmented control logic is executed: when When the observation is normal, the current observation noise covariance matrix is ​​maintained and a regular update is performed. when At that time, through the magnification factor Reduce the weight of anomalous observations in Kalman correction and generate an adaptively corrected observation noise covariance matrix. : 。 10. A USV-ROV multi-source fusion relative positioning system based on cable dynamic constraints, characterized in that, include: Unmanned surface vessels, cable management systems, and underwater remotely operated robots; A dual-antenna direction-finding real-time dynamic carrier phase differential module and a high-frequency inertial measurement unit are configured at the end of the unmanned surface vessel to realize state prediction and state correction, respectively. The outputs of the real-time dynamic carrier phase differential module and the high-frequency inertial measurement unit are fused to generate a continuous and drift-free global pose reference vector for the unmanned surface vessel. The cable management system includes a cable length measurement unit and a cable azimuth measurement unit, which are used to collect the cable length and cable azimuth, respectively; a depth gauge is used to collect the vertical depth difference relative to the water surface; By introducing dynamic equivalent parameters, the spatial positional relationship between the underwater remotely operated robot and the surface unmanned vessel is established, and the three-dimensional position observation vector of the underwater remotely operated robot in the global reference coordinate system is constructed. Based on the global pose reference vector of the unmanned surface vessel and the three-dimensional observation vector of the underwater remotely operated robot in the global navigation coordinate system, an adaptive observation noise covariance online adjustment strategy is introduced under the Kalman filter framework to generate the optimal state estimation vector and global position components of the underwater remotely operated robot. The cable-based dynamic constraint-based USV-ROV multi-source fusion relative positioning system is used to implement the cable-based dynamic constraint-based USV-ROV multi-source fusion relative positioning method and its steps as described in claim 1.