Multi-source sensing data fusion unmanned ship draft real-time detection and adjustment method and system

Abstract

The application discloses a multi-source sensing data fusion unmanned ship draft real-time detection and adjustment method and system, and relates to the technical field of industrial control.The method comprises the following steps: reading multi-source sensing data to construct a ship body state sequence; synchronizing the sequence, associating the attitude and the draft change, and extracting the response time lag; decomposing the state sequence according to the time lag to obtain transient disturbance and steady-state response; extracting the part satisfying the stable condition in the steady state to generate an evolution sequence by combining the historical disturbance characteristics; determining the target window based on the attitude stability and motion continuity of the evolution sequence; and using the window to configure the control strategy to execute the stable adjustment of the unmanned ship. The technical problems that the existing unmanned ship cannot accurately detect the draft change in real time and dynamically adjust the ship body attitude under complex sea conditions, resulting in low draft state sensing accuracy and the inability to perform stable adjustment in advance for unmanned aerial vehicle deployment and recovery operation requirements are solved, and the technical effects of improving the stability and landability of the unmanned ship under dynamic sea conditions are achieved.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology, specifically to a method and system for real-time draft detection and adjustment of unmanned vessels based on multi-source sensor data fusion. Background Technology

[0002] With the rapid development of the offshore wind power industry, unmanned surface vessels (USVs), as autonomous operating platforms, are widely used in tasks such as wind farm inspection, underwater structure inspection, and environmental monitoring. However, USVs constantly face complex and ever-changing surface environmental disturbances during offshore operations. The coupling effect of external forces such as wind, waves, and currents causes the hull to undergo six degrees of freedom of random motion, which not only affects the navigation safety and operational accuracy of the USV itself but also poses a severe challenge to the normal operation of onboard equipment. In unmanned wind power inspection scenarios, USVs need to carry drones to complete autonomous take-off and landing tasks. The success or failure of drone landing is highly dependent on the attitude stability and positional accuracy of the hull platform at the moment of landing. At the same time, the deployment and retrieval of underwater equipment such as ROVs also require the hull to have a stable draft and attitude control capabilities. Therefore, achieving real-time detection and dynamic adjustment of the USV's draft and hull attitude is a key technological prerequisite for ensuring the safety and reliability of unmanned offshore operations. In existing technologies, ship draft detection mainly relies on manual reading of draft marks or measurement by single-point level sensors, which suffers from drawbacks such as low measurement accuracy, poor real-time performance, and difficulty in adapting to dynamic sea conditions. In terms of unmanned surface vessel (USV) attitude control, although there are technical solutions for adjusting draft and longitudinal and lateral stability by pumping water into and out of ballast tanks, as well as dynamic positioning methods based on wave perception, these solutions often treat draft detection, attitude perception, and control adjustment as independent processes. They lack correlation analysis between changes in ship draft and attitude response under water surface disturbance conditions, resulting in a phase mismatch between adjustment actions and actual changes in the ship's state, making it difficult to achieve the expected control effect. Summary of the Invention

[0003] This application provides a method and system for real-time draft detection and adjustment of unmanned surface vessels (USVs) based on multi-source sensor data fusion. The key feature is that it solves the technical problem in the prior art that USVs cannot accurately detect draft changes in real time and dynamically adjust their hull attitude under complex sea conditions, resulting in low draft state perception accuracy and the inability to make advance stable adjustments for the needs of USV deployment and recovery operations. It achieves the technical effect of improving the stability and landing capability of USVs under dynamic sea conditions and ensuring operational safety by using multi-source sensor data fusion and analysis of the hydrodynamic response lag relationship between attitude changes and draft changes.

[0004] The first aspect of this application provides a method for real-time draft detection and adjustment of an unmanned surface vessel through multi-source sensor data fusion, the method comprising: The system reads multi-source sensor data from the unmanned surface vessel (USV), including draft measurement data, attitude measurement data, and surface disturbance perception data. A unified hull state representation sequence is constructed based on this data. The hull state representation sequence is then time-synchronized to construct attitude change and draft change sequences. Time correlation analysis is performed on these sequences to extract the time lag relationship between attitude change and draft change responses, generating hydrodynamic response lag parameters. These parameters are then used to decompose the hull state representation sequence, establishing transient disturbance components and steady-state response components. The portion of the steady-state response component that meets preset stability conditions is extracted as the current stable hull state. The hydrodynamic response lag parameters, combined with the variation characteristics of transient disturbance components within historical time windows, are used for time extrapolation to generate an evolution sequence of the hull stable state. Based on this evolution sequence, attitude stability constraints, and platform motion continuity constraints, a landing capability determination is performed, and a target stability window is constructed. Finally, a stability control strategy is configured using the target stability window to perform unmanned surface vessel stability adjustment.

[0005] A second aspect of this application provides a real-time draft detection and adjustment system for unmanned vessels based on multi-source sensor data fusion, the system comprising: Data Reading Unit: Reads multi-source sensor data from the unmanned surface vessel (USV), including draft measurement data, attitude measurement data, and surface disturbance perception data. Based on this multi-source sensor data, a unified hull state representation sequence is constructed. Time Correlation Analysis Unit: Performs time synchronization processing on the hull state representation sequence, constructs attitude change sequences and draft change sequences, performs time correlation analysis on these sequences, extracts the time lag relationship between attitude change responses and draft change responses, and generates hydrodynamic response lag parameters. Sequence Decomposition Unit: Decomposes the hull state using the hydrodynamic response lag parameters. The system employs a sequence decomposition method to establish transient disturbance components and steady-state response components. A time extrapolation unit extracts the portion of the steady-state response component that meets preset stability conditions as the current stable state of the hull. It then uses hydrodynamic response lag parameters combined with the variation characteristics of transient disturbance components within historical time windows to perform time extrapolation, generating an evolution sequence of the hull's stable state. A landing capability determination unit performs landing capability determination based on the evolution sequence, attitude stability constraints, and platform motion continuity constraints, constructing a target stability window. A stability adjustment unit configures a stability control strategy using the target stability window and performs unmanned surface vessel stability adjustment.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, sensor information on the unmanned surface vessel's (USV) draft, attitude, and surface disturbances is integrated to form time-series data that comprehensively reflects the vessel's operational status. Then, the various data types are time-aligned, and the response sequence between attitude changes and draft changes is analyzed to determine the hysteresis characteristics of the vessel under hydrodynamic forces. Next, short-term fluctuations and relatively stable changes in the vessel's state are separated to identify the current true stable state, and subsequent stable states are predicted based on historical disturbance trends. Then, based on the prediction results and requirements for attitude stability and motion continuity, a suitable stable time window for USV landing is determined. Finally, a control strategy is developed around this window to stabilize the USV, enabling it to maintain a stable state that meets operational requirements under complex sea conditions. Attached Figure Description

[0007] 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.

[0008] Figure 1 This is a schematic flowchart of a method for real-time draft detection and adjustment of unmanned vessels based on multi-source sensor data fusion, provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of a landingability determination confusion matrix provided in an embodiment of this application.

[0010] Figure 3 This is a schematic diagram of the structure of the unmanned vessel draft real-time detection and adjustment system based on multi-source sensor data fusion provided in this application embodiment.

[0011] Figure labeling: Data reading unit 11, time correlation analysis unit 12, sequence decomposition unit 13, time extrapolation unit 14, landing determination unit 15, stability adjustment unit 16. Detailed Implementation

[0012] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0013] Example 1, as Figure 1 As shown, this application provides a method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion, wherein the method includes: Read multi-source sensor data from the unmanned vessel, including draft measurement data, attitude measurement data, and surface disturbance perception data, and construct a unified hull state representation sequence based on the multi-source sensor data.

[0014] In this embodiment, draft measurement devices are installed on the port, starboard, port, and starboard sides of the bow of the unmanned vessel, an attitude measurement device is installed near the ship's center of gravity, and a surface disturbance sensing device is installed above the bow. The system operates at a preset main sampling frequency, reading the pressure value from the draft measurement device, the three-axis acceleration and three-axis angular velocity output from the attitude measurement device, and the set of water surface distance points in front of the bow output from the surface disturbance sensing device. For any draft measurement device, the current atmospheric pressure is first subtracted from the sensor pressure value, and then the liquid column height at the sensor location is calculated based on the water density and gravitational acceleration. Combined with the sensor's installation height relative to the ship's reference plane, the actual draft value at the corresponding measurement point is calculated. Subsequently, based on the draft values ​​at the measuring points, the average draft, bow-stern draft difference, and port-board draft difference are calculated to form draft measurement data. The average draft characterizes the overall hull sinking; the bow-stern draft difference characterizes the uneven draft caused by trim; and the port-board draft difference characterizes the uneven draft caused by list. For any attitude measurement device, the coordinates are transformed to the hull coordinate system according to a pre-calibrated installation matrix, and attitude calculations are used to obtain the roll angle, pitch angle, roll angular velocity, pitch angular velocity, vertical acceleration, and heave displacement, forming attitude measurement data. For the surface disturbance sensing device, the average water surface height for each distance interval is calculated to form a wave height sequence. The current wave height, wave period, and main wave direction are calculated from the continuous wave height changes, forming surface disturbance sensing data to characterize the surface disturbance state ahead of the hull. Each time data is read, the sensor number, source sampling time, controller reception time, and validity identifier are written into the data. When the pressure value exceeds the range, the attitude measurement data shows discontinuous jumps, or the number of effective surface points is lower than the set value, the corresponding data is marked as invalid, but its time identifier is retained, thus forming multi-source sensor data. After the above processing is completed, the current draft measurement data, attitude measurement data, and surface disturbance sensing data are combined into a hull state vector according to the preset field order. Then, multiple consecutive hull state vectors form a unified hull state representation sequence, providing input for subsequent time correlation analysis of attitude changes and draft changes.

[0015] Furthermore, the acquisition of the water surface disturbance sensing data includes at least one of the following methods: a: By using millimeter-wave radar, lidar, or monocular / binocular vision sensors installed on the hull, collect wave height sequences or wave image sequences around the ship, and extract wave height, wave period, and wave direction parameters as water surface disturbance sensing data; b: By using the shipborne inertial wave sensing module, obtain wave statistical parameters at the ship's location based on the vertical acceleration and angular velocity sequence inversion of a six-degree-of-freedom IMU as water surface disturbance sensing data; Align the water surface disturbance sensing data with the draft measurement data and attitude measurement data on the time axis to construct the ship's state characterization sequence.

[0016] Preferably, after installing a surface disturbance sensing device, a draft measurement device, and an attitude measurement device on the unmanned surface vessel, a unified clock source is configured for each sensor. The draft measurement device may include pressure-type draft sensors arranged on the port bow, starboard bow, port stern, and starboard stern; the attitude measurement device may be a six-degree-of-freedom IMU; the surface disturbance sensing data may include millimeter-wave radar, lidar, monocular / binocular vision sensors, and shipborne inertial wave sensing modules. Specifically, when using millimeter-wave radar, lidar, or monocular / binocular vision sensors installed near the bow, superstructure, or UAV landing platform, a predetermined water surface area around the ship will be continuously scanned. Taking millimeter-wave radar as an example, the water surface in front of the bow is scanned according to a predetermined sampling frequency, and at each sampling moment, the radar detection point is output, including the target distance, azimuth angle, pitch angle, and echo intensity. The system calculates the three-dimensional coordinates of each detection point in the radar coordinate system based on its distance and angle. Using a pre-calibrated rotation matrix and translation vector, it transforms the radar detection points to the ship's coordinate system. The surface detection points are then divided into multiple surface grids at equal intervals along the ship's hull. The median height of the effective detection points within each grid is calculated to obtain the instantaneous wavefront height for each grid, forming multiple wavefront height sequences. Next, the average value within a sliding time window is subtracted from each wavefront height sequence to remove low-frequency offsets caused by tidal changes and overall ship heave. Bandpass filtering is then applied to the mean-free sequence to obtain the effective wavefront undulation sequence, which serves as the ship's perimeter water surface wavefront height sequence. Subsequently, according to a preset wave analysis window (e.g., 20 seconds), the standard deviation of the wavefront undulation sequence is calculated, and a preset multiple of the standard deviation is determined as the effective wave height. A fast Fourier transform is performed on the ship's perimeter water surface wavefront height sequence to obtain the wavefront power spectrum. The frequency corresponding to the peak value of the power spectrum is determined as the dominant wave frequency, and the wave period is calculated based on the reciprocal of the dominant wave frequency. For multiple virtual sampling points with different spatial locations, the cross-correlation function between their wave surface undulation sequences is calculated to determine the time difference of the same wave peak reaching different sampling points. Then, combined with the positional relationship of each sampling point in the ship's coordinate system, the wave propagation direction is calculated, and the wave direction angle relative to the bow direction is obtained as the wave direction parameter. The system combines the effective wave height, wave period, and wave direction parameters obtained within the current analysis window into a set of water surface disturbance sensing data. If a lidar is used, the acquired 3D point cloud of the water surface is divided into multiple regular grids. The median height of the point cloud in each grid is calculated to obtain the spatiotemporal variation sequence of the water surface waveform. Multiple fixed grids are then selected from the spatiotemporal variation sequence of the water surface waveform, and the wave surface height sequence of each grid is extracted. The effective wave height is calculated based on the standard deviation of the wave surface height sequence, the wave period is calculated based on the peak power spectrum, and the wave propagation speed and wave direction angle are calculated based on the spatial displacement and time difference of the same wave peak in adjacent grids.When using monocular / binocular vision sensors, wave image sequences are acquired, and the image coordinates are converted into the water surface spatial position in the ship's coordinate system to obtain the wave surface height estimate at each sampling time, forming a wave surface height sequence. The wave height is then calculated based on the height difference between the wave crest and the wave trough, the wave period is calculated based on the time interval between adjacent wave crests passing the same observation position, and the wave direction angle is calculated based on the movement direction of the wave crest features in the continuous images.

[0017] Furthermore, a six-degree-of-freedom (6DOF) IMU can be used to acquire the motion response of the hull after being subjected to waves, and the wave statistical parameters at the hull's location can be retrieved based on the correspondence between the hull's motion response and wave excitation. Specifically, the 6DOF IMU is installed near the center of gravity of the unmanned surface vessel (USV), and its position and installation orientation relative to the hull coordinate system are pre-calibrated. The 6DOF IMU outputs three-axis acceleration, three-axis angular velocity, and attitude quaternions in the hull coordinate system according to a preset sampling frequency. The system converts the three-axis acceleration in the hull coordinate system to the geographic coordinate system based on the attitude quaternions, and subtracts the gravitational acceleration from the vertical acceleration to obtain the hull's vertical motion acceleration. Then, the vertical acceleration is bandpass filtered to retain the signal within the main frequency range corresponding to the ocean waves. The filtered vertical acceleration is integrated for the first time to obtain the heave displacement, and then integrated for the second time to obtain the hull's heave displacement. During the integration process, the error between the ship's motion responses obtained through high-pass filtering is used. The candidate wave direction with the smallest response error is determined as the current wave direction, and the wave energy spectrum corresponding to this candidate wave direction is determined as the inverted wave spectrum. The system integrates the inverted wave spectrum to obtain the zero-order spectral moment, calculates the significant wave height based on the zero-order spectral moment, calculates the wave period based on the spectral peak frequency in the inverted wave spectrum, and uses the candidate wave direction with the smallest error as the wave direction statistical parameter. Thus, surface disturbance sensing data including significant wave height, wave period, and wave direction are obtained. After acquiring the surface disturbance sensing data, it is aligned with the draft measurement data and attitude measurement data on the time axis. That is, a unified target time series is established at a preset time interval, and the corrected sampling time corresponding to each data is calculated based on the source sampling timestamp of each sensor and the pre-calibrated fixed acquisition delay. For example, two effective sampling values ​​located before and after the target time are selected from the draft measurement data and linearly interpolated to obtain the draft value of each measuring point at the target time. Finally, according to the preset field order, the draft measurement data, attitude measurement data, and surface disturbance sensing data of each measuring point at the target time are combined into a hull state vector. The continuously generated hull state vectors are then arranged sequentially according to the target time series to construct a hull state representation sequence, providing a data foundation for the subsequent time correlation analysis of attitude change response and draft change response.

[0018] The ship's state characterization sequence is time-synchronized to construct attitude change sequence and draft change sequence. Time correlation analysis is performed on the attitude change sequence and draft change sequence to extract the time lag relationship between attitude change response and draft change response, and hydrodynamic response lag parameters are generated.

[0019] In one embodiment, the hull state representation sequence is rearranged onto the target time axis at a preset uniform processing frequency, such as 20Hz, and the source sampling timestamps of each sensor are used as the alignment basis. For data where the time interval between two adjacent valid sampling points does not exceed a preset valid interval, linear interpolation is performed according to the proportion of the target time between the two sampling times to obtain the draft data, attitude data, and surface disturbance data corresponding to each time point. Subsequently, using the most recent time interval, such as 30 seconds, as a sliding analysis window, the roll angle, pitch angle, and heave displacement are extracted from the synchronized attitude data. The changes at adjacent sampling times are calculated to obtain the roll change sequence, pitch change sequence, and heave change sequence. The roll change at the k-th time is the difference between the roll angle at that time and the roll angle at the previous time. The pitch change and heave change are calculated in the same way. At the same time, the difference in draft between port and starboard sides, the difference in draft between bow and stern, and the average draft are calculated based on the data from the draft measurement points. The changes in the difference in draft between port and starboard sides at adjacent times are used as the draft change sequence corresponding to roll, the changes in draft between bow and stern sides at adjacent times are used as the draft change sequence corresponding to pitch, and the changes in the average draft at adjacent times are used as the draft change sequence corresponding to heave. This forms three sets of mutually corresponding attitude change sequences and draft change sequences: roll-port and starboard draft, pitch-bow and stern draft, and heave-average draft. To eliminate the influence of differences in the dimensions and amplitudes of different physical quantities, the mean and standard deviation of each sequence within the current analysis window are calculated. The mean is subtracted from each sample value, and then the result is divided by the standard deviation to obtain the standardized attitude change sequence and the standardized draft change sequence. For each set of sequences, the draft change sequence is moved point by point within the candidate hysteresis range, moving one sampling point at a time, and the normalized cross-correlation coefficient between the attitude change sequence and the moved draft change sequence is calculated. The system determines the movement with the largest absolute value of the cross-correlation coefficient as the candidate hysteresis point number for that set of responses. The candidate hysteresis point number is multiplied by a preset sampling interval to obtain the corresponding time hysteresis value, and the sign of the largest cross-correlation coefficient is retained to represent the same-direction or opposite-direction response relationship between attitude change and draft change. To prevent erroneous matches caused by random fluctuations, the corresponding time lag value is only considered valid if the absolute value of the maximum cross-correlation coefficient is not less than the lag correlation determination threshold (e.g., not less than 0.4), and the maximum value has a peak difference with the cross-correlation coefficient of the adjacent lag position that is not less than the correlation peak significance threshold (e.g., a peak difference of not less than 0.05). Otherwise, the valid time lag value obtained from the previous analysis window is used, and the confidence level of the current result is marked as low.After obtaining the roll response lag time, pitch response lag time, and heave response lag time respectively according to the above process, the absolute values ​​of the three sets of maximum cross-correlation coefficients are used as weights to perform a weighted average of the three sets of effective time lag values, yielding the comprehensive hydrodynamic response lag time of the hull within the current analysis window. The roll response lag time, pitch response lag time, heave response lag time, comprehensive hydrodynamic response lag time, each set of cross-correlation coefficients, and their corresponding confidence levels are then combined to form the hydrodynamic response lag parameter, used to characterize the degree of delay in the draft measurement response after a change in hull attitude. Through this processing, the response delay from attitude change to draft change can be accurately characterized, improving the accuracy of subsequent disturbance decomposition and steady-state identification.

[0020] The hull state characterization sequence is decomposed using hydrodynamic response hysteresis parameters to establish transient disturbance components and steady-state response components.

[0021] In one embodiment, let the hull state vector at the k-th target time be X(k), and read the roll response lag time τ generated in the previous processing step. r Pitch response lag time τ p Heave response lag time τ h and the overall hydrodynamic response lag time τ c And according to a uniform sampling interval Δt, each lag time is converted into the corresponding number of lag sampling points, where N r =round(τ) r / Δt), N p =round(τ) p / Δt), N h =round(τ) h / Δt), N c =round(τ) c / Δt). To ensure that the disturbance state that causes the ship's motion corresponds to the subsequent draft response at the same time, the port and starboard draft difference at the k-th time is compared with the draft response at the (k-N)-th time, using the time position of the draft data as a reference. r The combination of the roll angle and roll rate at time k, and the bow-stern draft difference at time k-N, is used to... p The combination of the pitch angle and pitch velocity at time k, and the average draft at time k-N, is used to determine the average draft at time k-N. h The heave displacement and vertical acceleration at time k are combined, and the k-Nth time is... cThe effective wave height, wave period, and wave direction angle at each moment are used as the surface disturbance state corresponding to the draft response at the k-th moment, thus forming the hull state vector Z(k) after hysteresis compensation. When k is less than the corresponding number of hysteresis sampling points, the decomposition result is not output until the data length accumulated in the buffer reaches the maximum number of hysteresis sampling points. Subsequently, the steady-state extraction time constant is set according to the hydrodynamic response hysteresis parameters, and T is used for the average draft. s,h =K h τ h T-type water difference between port and starboard sides and roll condition are used. s,r =K r τ r T was used to assess the difference in draft between bow and stern and the pitching condition. s,p =K p τ p T is used to assess the disturbance state of the water surface. s,c =K c τ c , where K h K r K p K c These are the multiples of the hydrodynamic response lag times preset according to the ship's response characteristics. The system determines the corresponding time constant T according to the response relationship of each state field. j And calculate the recursive coefficient α of this field. j =Δt / (T) j +Δt). For the j-th state field, according to S j (k)=S j (k-1)+α j [Z j (k)-S j (k-1)] is recursively calculated, and S j (k) is determined as the steady-state response value of this field at time k, and according to D j (k)=Z j (k)-S j (k) Calculate the corresponding transient disturbance value. Upon initial decomposition, read the valid state data from the previous preset time period and calculate the median of each field as the initial steady-state response value S. j (0), thus avoiding abrupt changes in decomposition results caused by initial single-point anomalies. After each target time step calculation is completed, the S values ​​of each field are... j (k) Combine the fields of the original hull state vector into steady-state response components S(k), and then combine the D values ​​of each field. jThe components (k) are combined to form the transient disturbance component D(k), satisfying Z(k) = S(k) + D(k). The steady-state response component sequence is constructed from the continuous time intervals of S(k), and the transient disturbance component sequence is constructed from the continuous time intervals of D(k). Thus, the corresponding positions and decomposition time scales of the disturbance and draft response are determined by utilizing the measured hydrodynamic response lag, reducing the interference of short-term measurement fluctuations caused by waves on the identification of the actual draft and the ship's stability state.

[0022] The portion of the steady-state response component that meets the preset stability condition is extracted as the current stable state of the hull. The evolution sequence of the stable state of the hull is generated by extrapolating the time using the hydrodynamic response lag parameter combined with the variation characteristics of the transient disturbance component in the historical time window.

[0023] In one embodiment, the continuously generated steady-state response component sequence S(k) and transient disturbance component sequence D(k) are used as inputs, and the hull stability state is identified and extrapolated using a uniform processing frequency. For each stability determination window, the range and standard deviation of the mean draft, the absolute maximum value of the bow-stern draft difference, the absolute maximum value of the port-ship draft difference, the absolute maximum value of the roll angle, the absolute maximum value of the pitch angle, the root mean square value of the roll angular velocity, the root mean square value of the pitch angular velocity, and the proportion of effective data are calculated sequentially within that window. When the average draft range is not greater than the average draft fluctuation range threshold, the average draft standard deviation is not greater than the average draft dispersion threshold, the absolute maximum value of the bow-stern draft difference is not greater than the longitudinal draft difference threshold, the absolute maximum value of the port-starboard draft difference is not greater than the transverse draft difference threshold, the absolute maximum value of the roll angle is not greater than the roll angle stability threshold, the absolute maximum value of the pitch angle is not greater than the pitch angle stability threshold, the root mean square value of the roll angular velocity is not greater than the roll angular velocity fluctuation threshold, the root mean square value of the pitch angular velocity is not greater than the pitch angular velocity fluctuation threshold, and the proportion of valid data is not lower than the proportion of valid data threshold, the judgment window is marked as a stable window. After confirming the stable state, the median of each steady-state response field within the continuous stable interval is calculated. The obtained median average draft, median bow-stern draft difference, median port-starboard draft difference, median roll angle, median pitch angle, median heave displacement, and median angular velocity are combined to form the current hull stability state B(k), and the start time, duration, and proportion of valid data of the stable state are recorded. If the stability conditions are not met at the current moment, the most recent valid hull stability state is retained, and its confidence level is reduced according to a preset rule, for example, by 10% per second. If no new valid stability state is obtained for 10 consecutive seconds, the current hull stability state is marked as invalid.

[0024] When generating the evolution sequence of the ship's stable state, τ is read. r τ p τ h and τ c And convert them to N respectively.r N p N h and N c Using the transient disturbance components within a preset basic historical window duration as the historical time window, when the overall hydrodynamic response lag time exceeds the preset window expansion trigger duration, the historical time window is expanded to the product of the overall hydrodynamic response lag time and the preset lag window expansion multiple, ensuring that the expanded historical time window does not exceed the preset historical window duration upper limit. For average draft transient disturbances, port / port draft difference transient disturbances, bow / stern draft difference transient disturbances, roll angle transient disturbances, and pitch angle transient disturbances, second-order autoregressive prediction relationships are established respectively; that is, the next sampled value is predicted using the two transient disturbance values ​​preceding the current sampled value. Specifically, the first autoregressive coefficient 'a' is obtained using the least squares method within the historical time window. 1,j Second autoregression coefficient a 2,j To minimize the sum of squared prediction errors, and according to D̂ j (k+q)=a 1,j D̂ j (k+q-1)+a 2,j The predicted value of the future transient disturbance is obtained by recursively calculating D̂j(k+q-2), where q represents the number of prediction steps. The initial values ​​for the recursion are the actual transient disturbance values ​​at the current and previous times. To determine the degree of influence of the transient disturbance on the subsequent steady state, the n-Nth time step is calculated within the historical time window. j The transient disturbance value D at each moment j (n-N j ) and the steady-state response increment ΔS at time n j (n)=S j (n)-S j Pairing is performed using (n-1) pairs, and according to γ j =Σ[D j (n-N j )ΔS j (n)] / Σ[D j (n-N j )] 2 Calculate the disturbance propagation coefficient γ of the corresponding state field. j Among them, the average draft is measured using N h The difference in draft between port and starboard sides and the roll angle are determined using N. r The draft difference between bow and stern and the pitch angle are calculated using N. p When the denominator is less than the preset minimum energy threshold, the corresponding perturbation transfer coefficient is set to zero to avoid numerical amplification when there is no obvious perturbation.

[0025] The system typically sets the prediction duration to three times the overall hydrodynamic response lag time and generates the future hull state point by point according to a preset sampling interval. For the j-th state field, the state value at the first prediction time is taken as the current stable hull state B. j (k), and subsequent prediction times are based on B̂ j (k+q)=B̂ j (k+q-1)+γ j D j * (k+q-N j Perform recursive calculations. When k+q-N j No later than the current time, D j * Using the actual transient disturbance values ​​at the corresponding historical moments, when k+q-N j Later than the current time, D j * Transient disturbance predictions are generated using a second-order autoregressive model. After performing the aforementioned recursion on mean draft, bow-stern draft difference, port-ship draft difference, roll angle, pitch angle, and heave displacement, the fields at the same prediction time are combined into a predicted hull stability vector B̂(k+q), and arranged chronologically to form an evolution sequence of hull stability states. Simultaneously, the prediction residuals of the autoregressive model within the historical time window are calculated. When the root mean square value of the residual exceeds a preset proportion of the standard deviation of the corresponding transient disturbance, the prediction confidence of that field is reduced; when the prediction confidence is below a preset threshold, the effective prediction time of that field is shortened, and results exceeding the effective prediction time are not used for subsequent landing suitability determination. Through this processing, the current hull state with continuous stability characteristics can be extracted from the steady-state response, and combined with actual hydrodynamic lag and historical disturbance change trends, the subsequent hull state can be predicted, providing a reliable basis for early identification of landing-ready stable periods.

[0026] Based on the evolution sequence, attitude stability constraints, and platform motion continuity constraints, a landing capability determination is performed, and a target stability window is constructed.

[0027] In one embodiment, the hull stability state evolution sequence generated at a preset sampling interval is used as input. Each predicted state vector in the evolution sequence includes at least the predicted average draft, predicted bow-stern draft difference, predicted port-to-port draft difference, predicted roll angle, predicted pitch angle, predicted heave displacement, predicted roll angular velocity, predicted pitch angular velocity, and corresponding prediction confidence level. The system first sets attitude stability constraints according to the requirements of UAV landing operations. For example, the absolute value of the predicted roll angle is no greater than 2°, the absolute value of the predicted pitch angle is no greater than 1.5°, the absolute value of the predicted roll angular velocity is no greater than 1° / s, the absolute value of the predicted pitch angular velocity is no greater than 0.8° / s, the difference between the predicted average draft and the average draft in the current stable hull state is no greater than 0.03m, the absolute value of the predicted bow-stern draft difference is no greater than 0.05m, the absolute value of the predicted port-to-port draft difference is no greater than 0.03m, and the confidence level of the predicted state is no less than 0.7. For the q-th predicted time in the evolution sequence, the corresponding predicted state is compared with the aforementioned thresholds. If all attitude stability constraints are satisfied, the attitude stability flag for that predicted time is set to 1; otherwise, it is set to 0. Subsequently, the continuity of platform motion between adjacent predicted times is determined. The changes in roll angle, pitch angle, heave displacement, mean draft, roll velocity, and pitch velocity between the q-th and (q-1)-th predicted times are calculated and divided by a preset sampling interval to obtain the corresponding rate of change. When the changes in roll angle, pitch angle, heave displacement, mean draft, roll velocity rate of change, and pitch velocity rate of change for adjacent predicted times are all less than the corresponding rate of change thresholds, the adjacent time period is determined to satisfy the platform motion continuity constraints. To further eliminate spurious stable states caused by frequent reversals in the ship's motion direction, roll and pitch angular velocities are continuously read at multiple prediction times. If the sign of the angular velocity reverses twice within two adjacent sampling intervals, or if the rate of change of the angular velocity exceeds the corresponding continuity threshold, the platform motion continuity flag at the intermediate prediction time is set to 0. In other cases, the platform motion continuity flag is set to 1.

[0028] The system performs a logical AND operation between the attitude stability flag and the platform motion continuity flag at each predicted moment to obtain a landable flag sequence. When both flags at a given predicted moment are 1, that moment is marked as a landable moment. The system then scans the landable flag sequence chronologically, merging adjacent and consecutive predicted moments of 1 into candidate stability intervals, and calculates the start and end times and duration of each candidate stability interval. Based on the minimum stabilization time required for the UAV to complete approach, descent, and contact with the deck, the system retains only candidate stability intervals with a duration not less than the minimum stabilization time. Safety margins are deducted from the start and end of each candidate stability interval to prevent the UAV from landing just as the hull enters or is about to exit a stable state. Intervals whose duration after deducting the safety margin is still not less than the minimum duration threshold of the effective stability window are determined as effective stability windows. For each effective stabilization window, the maximum absolute value and standard deviation of the predicted roll angle, predicted pitch angle, predicted heave displacement, and predicted mean draft within the window are calculated. The average reliability is calculated based on the reliability of each prediction time. The standard deviations of the roll angle, pitch angle, heave displacement, and mean draft are divided by their respective allowable thresholds and summed to obtain the window motion fluctuation. The stability window evaluation value is then obtained by dividing the product of the window duration and the average reliability by the window motion fluctuation value. The system prioritizes selecting the effective stabilization window with the largest evaluation value, whose start time is no more than the maximum waiting time threshold of the target window, whose average reliability is not lower than the average reliability threshold of the target window, and whose start time is no more than the maximum waiting time threshold of the target window. When no effective stabilization window that meets the minimum duration requirement exists in the evolution sequence, a current no-landing flag is output, and the system waits for the next update of the ship's stability state evolution sequence before re-performing the determination. After the target stabilization window is determined, the expected start time, expected end time, target mean draft, and target roll angle are recorded and used as the basis for subsequent stability control strategy configuration and UAV landing command generation.

[0029] The target stability window is used to configure the stability control strategy and execute the stability adjustment of the unmanned vessel.

[0030] In one embodiment, data such as the target average draft, target roll angle, target pitch angle, target heave displacement, and allowable fluctuation range corresponding to the target stability window are read and compared with the current hull stability state. The existing state deviation is calculated, and the hull stability control quantity is configured as a stability control strategy according to the magnitude and direction of each state deviation. The configured stability control strategy is then sent to the ballast adjustment unit, propulsion control unit, and attitude adjustment unit respectively to collaboratively correct the hull draft, attitude, and motion state, so that the unmanned vessel can achieve stability requirements within the target stability window.

[0031] Furthermore, by utilizing the target stability window to configure a stability control strategy and performing unmanned surface vessel (USV) stability adjustments, including: Read the state deviation between the target stability window and the current hull stability state; configure the hull stability control quantity according to the state deviation, and send the hull stability control quantity to the ballast adjustment unit, propulsion control unit and attitude adjustment unit to perform coordinated adjustment.

[0032] Preferably, after the target stability window is determined, the start time, end time and predicted hull state within the window are read first, and the median of the target average draft, target bow-stern draft difference, target port-ship draft difference, target roll angle, target pitch angle, target heave displacement, target roll angular velocity and target pitch angular velocity within the window are calculated respectively. The above medians are combined into a target state vector. Subsequently, the most recent valid current hull stability state is read, and the mean draft deviation, bow / stern draft deviation, port / board draft deviation, roll angle deviation, pitch angle deviation, heave displacement deviation, roll rate deviation, and pitch rate deviation are calculated item by item according to the target state value minus the current state value. This forms a state deviation vector, where a positive mean draft deviation indicates that the hull draft needs to be increased; a positive bow / stern draft deviation indicates that the bow draft needs to be increased relatively; and a positive port / board draft deviation indicates that the starboard draft needs to be increased relatively. The positive and negative directions of the attitude angle and angular velocity deviations are pre-defined according to the hull coordinate system. To avoid frequent actions of the actuators caused by small sensor fluctuations, the controller sets dead zones for mean draft deviation, bow / stern and port / board draft deviation, roll and pitch angle deviation, and angular velocity deviation. When the absolute value of a deviation is less than the corresponding dead zone, the deviation is set to 0. After dead zone processing, the maximum permissible deviation of each state variable is used as the normalization benchmark to convert the state deviation vector into a dimensionless deviation vector. This vector is then used to construct a characterization of the coupling relationship between draft deviation, attitude deviation, and motion deviation, describing the direction and intensity of mutual influence among the deviation components. Subsequently, this coupling relationship is corrected using hydrodynamic response hysteresis parameters to obtain the coupling influence weights of each deviation component at different time scales. This weight is then used for collaborative constraint processing of the initial control demand set, suppressing control conflicts that may arise from coupling effects among the execution units, thereby constructing the hull stability control variables used to define the stability control strategy. Then, the target flow rates corresponding to the bow, stern, port, and starboard ballast tanks in the stability control strategy are converted into ballast pump speeds and valve openings. The port and starboard propulsion control requirements are converted into target thrust for the port and starboard propellers, and the attitude adjustment requirements are converted into target angles for the anti-roll fins. These are then distributed to the ballast adjustment unit, propulsion control unit, and attitude adjustment unit, respectively. Each execution unit returns the actual flow rate, actual thrust, actual anti-roll fin angle, and execution completion status. The system rereads the hull state and updates the state deviation based on the feedback results. When the average draft deviation is not greater than the allowable threshold for average draft deviation, the difference between the bow and stern drafts and the difference between the port and starboard drafts are not greater than the allowable thresholds for the difference between the bow and stern drafts and the allowable thresholds for the difference between the port and starboard drafts, the roll angle is not greater than the allowable threshold for the roll angle, the pitch angle is not greater than the allowable threshold for the pitch angle, and the corresponding angular velocity remains within the allowable range, the control input is stopped from increasing and the system is switched to the stability holding mode. The ballast valves are kept closed within the target stability window, and transient disturbances are suppressed only by slightly adjusting the thruster thrust and the anti-roll fin angle.Through the above processing, the stability requirements corresponding to the target stability window can be transformed into control commands that can be directly executed by each execution unit, and the ballast, propulsion and attitude adjustment actions can be coordinated with each other, so that the unmanned vessel can enter and maintain the predetermined stability state before the target stability window is reached.

[0033] Furthermore, configuring hull stability control quantities based on the aforementioned state deviation includes: Based on the state deviation, a coupling relationship representation is constructed between draft deviation, attitude deviation, and motion deviation. This coupling relationship representation describes the direction and intensity of mutual influence between each deviation component. The coupling relationship representation is then time-series corrected using the hydrodynamic response hysteresis parameter to obtain the coupling influence weights of each deviation component at different time scales. Based on these coupling influence weights, the state deviation is mapped to an initial control requirement set for multiple execution units, where the control requirement corresponding to any execution unit simultaneously includes the adjustment effect on multiple deviation components. The initial control requirement set is then subjected to collaborative constraint processing to suppress control conflicts caused by coupling effects between different execution units and to redistribute the control contributions of each execution unit, generating a consistent control requirement set. Based on the consistent control requirement set, hull stability control quantities are constructed and distributed to the ballast adjustment unit, propulsion control unit, and attitude adjustment unit for collaborative adjustment.

[0034] Optionally, using the state deviation between the target state vector corresponding to the target stability window and the current hull stability state vector as input, the system first uses the maximum permissible deviation corresponding to each deviation component as a normalization benchmark to convert each deviation into a dimensionless deviation ranging from -1 to 1. Then, it retrieves the state deviation change data within the most recent time period and forms a deviation history sequence according to a preset sampling interval. For any two deviation components, the value of one deviation component at the previous moment is used as input, and the change of the other deviation component at the next moment is used as output. The coupling coefficient between the two is calculated using least squares. Specifically, the product of the difference between the next deviation component and the previous deviation component is accumulated within the historical time window and divided by the accumulated result of the square of the previous deviation component to obtain the corresponding coupling coefficient. When the denominator is less than a preset minimum energy threshold, the coupling coefficient is set to zero. After calculating all deviation components pairwise, a coupling matrix is ​​formed. In this matrix, rows represent the affected deviation components, columns represent the contributing deviation components, the sign of the matrix elements indicates the direction of influence, and the absolute value of the elements indicates the intensity of influence. For example, if an increase in roll angle deviation leads to an increase in the difference in draft between the port and starboard sides in the same direction, the corresponding coupling coefficient is positive; if an increase in roll angle deviation leads to a decrease in another deviation, the corresponding coupling coefficient is negative. After characterizing the coupling relationships, the roll response lag time, pitch response lag time, heave response lag time, and overall hydrodynamic response lag time are read. Roll-related deviations are mapped to the roll response lag time, pitch-related deviations to the pitch response lag time, and mean draft and heave-related deviations to the heave response lag time. The system acquires a pre-set fast-change control time scale (e.g., 0.5s) and a pre-set slow-change control time scale (e.g., 5s), and performs time-series correction on each coupling coefficient according to the first-order response ratio. For the i-th deviation component, the coupling coefficient affected by the j-th deviation component is calculated using the fast-varying coupling influence weight. This weight is the product of the coupling coefficient and 1 minus the negative exponential value obtained by dividing the fast-varying control time scale by the corresponding hydrodynamic response lag time. The slow-varying coupling influence weight is calculated using the same method with the slow-varying control time scale. When the corresponding hydrodynamic response lag time is less than the preset sampling interval, it is processed according to the preset sampling interval. Therefore, deviation components with larger response lags receive smaller weights under the fast-varying control time scale and larger weights under the slow-varying control time scale, while deviation components with faster responses receive effective weights under the fast-varying control time scale. The system multiplies the fast-varying coupling influence weight matrix and the slow-varying coupling influence weight matrix with the current state deviation vector to obtain the fast-varying coupling compensation deviation and the slow-varying coupling compensation deviation, respectively. The corresponding compensation deviation is then added to the original state deviation to form the corrected state deviation used for control allocation by the execution unit.

[0035] The system pre-establishes an influence matrix for the execution units through static water tests. These execution units include ballast adjustment channels corresponding to the bow, stern, port, and starboard ballast tanks; propulsion control channels corresponding to the port and starboard vectoring jet pumps; and attitude adjustment channels corresponding to the port and starboard electric anti-roll fins. During calibration, step commands with preset proportional rated control values ​​are applied to each execution channel. Changes in average draft, bow-stern draft difference, port-starboard draft difference, roll angle, pitch angle, heave rate, roll rate, and pitch rate are recorded. The changes in each deviation component caused by a unit control value are then written into the execution unit influence matrix. The system primarily allocates the average draft deviation, bow-stern draft difference deviation, and port-side draft difference deviation in the slowly varying state deviation to the four ballast adjustment channels. It primarily allocates the roll angle deviation, pitch angle deviation, heave rate deviation, and angular velocity deviation in the rapidly varying state deviation to the port-side vectoring jet pumps and port-side electric anti-roll fins. However, each column of the actuator influence matrix retains the actual influence of that actuator on multiple deviation components. Therefore, the control requirement of any actuator simultaneously includes the adjustment effect on multiple deviation components. The system aims to approximate the state deviation correction by multiplying the actuator influence matrix and the control requirement vector, and uses constrained least squares to calculate the initial control requirement set. During calculation, a control quantity penalty term with preset coefficients is added to the sum of squared errors, and constraints are imposed on the ballast pump flow rate not exceeding the corresponding preset proportion of the rated flow rate, the vectoring jet pump thrust not exceeding the corresponding preset proportion of the rated thrust, and the electric anti-roll fin angle not exceeding the corresponding preset proportion of the rated angle, resulting in four ballast adjustment requirements, two propulsion control requirements, and two attitude adjustment requirements. Subsequently, based on the influence matrix of the execution unit, the contribution value of each initial control requirement to each deviation component is calculated. The system takes the direction of reducing the current deviation as the target action direction. When a contribution value is consistent with the target action direction, it is identified as a valid contribution; when a contribution value is opposite to the target action direction, it is identified as a conflict contribution. For cases where the same deviation component has reverse cancellation, its control conflict characteristics are determined, and these characteristics are used to adjust the initial control requirements of each execution unit in a coordinated manner. Then, the adjusted control requirements are normalized and redistributed to obtain a set of consistent control requirements that satisfy the conditions for coordinated convergence of multiple deviation components. Afterward, based on the set of consistent control requirements, hull stability control quantities are generated. The four ballast adjustment requirements are converted into the target flow rates and corresponding valve openings of the bow, stern, port, and starboard ballast pumps, respectively. The two propulsion control requirements are converted into the target thrust and nozzle direction of the port and starboard vector jet pumps, and the two attitude adjustment requirements are converted into the target angles of the port and starboard electric anti-roll fins.The system fills in the aforementioned control quantities according to a preset template to generate a stable control strategy, which is then distributed to the ballast adjustment unit, propulsion control unit, and attitude adjustment unit. It also reads the actual flow rate, actual thrust, actual nozzle direction, and actual anti-roll fin angle from each unit. In the next control cycle, it recalculates the state deviation and consistency control requirements. The average draft, bow-stern draft difference, and port-ship draft difference adjustment results output by the ballast adjustment unit serve as the reference state for the rapid-change control loop. The propulsion control unit and attitude adjustment unit only compensate for short-term attitude and motion disturbances in the reference state. Through this processing, considering the coupling and response lag between draft, attitude, and motion deviations, it can suppress the reverse cancellation and excessive superposition between different actuators, enabling the control actions of each actuator to jointly promote the coordinated convergence of multiple state deviations.

[0036] Furthermore, the initial set of control requirements is subjected to coordinated constraint processing, including: Based on the aforementioned coupling influence weights, a mapping relationship is constructed to show the influence of each execution unit's control action on draft deviation, attitude deviation, and motion deviation. The initial control requirements of each execution unit are projected onto the aforementioned influence mapping relationship to obtain the contribution distribution of each execution unit to each deviation component. Based on the contribution distribution of each execution unit on the same deviation component, the synergistic and counteracting effects between different execution units on that deviation component are identified, and control conflict characteristics are determined. The initial control requirements of each execution unit are adjusted in a coordinated manner according to the control conflict characteristics. The adjusted control requirements are then normalized and redistributed to generate a set of consistent control requirements that satisfy the conditions for coordinated convergence of multiple deviation components.

[0037] Optionally, taking the coupling influence weight matrix, the execution unit influence matrix, and the initial control requirement set as input, the system reads the current coupling influence weight matrix W and, according to H... ij =G ij +ΣW il G lj Calculate the corrected influence coefficient, where W il G represents the coupling influence weight of the l-th deviation component on the i-th deviation component. lj The direct influence of the j-th execution unit on the l-th deviation component is represented by the current influence mapping matrix H, composed of all corrected influence coefficients. Therefore, the j-th execution unit retains not only its direct control over each deviation component but also the indirect control effect resulting from this direct effect transmitted through the deviation coupling relationship. For the j-th initial control requirement u in the initial control requirement set... 0,j According to C ij =H ij u 0,jCalculate its projected contribution on the i-th deviation component, and form the contribution vector of the execution unit by combining all projected contributions corresponding to the same execution unit. Arrange the contribution vectors of all execution units in columns to form a contribution distribution matrix C. The system determines the current state deviation e. i Determine the target correction direction for the i-th deviation component, when e i When e is greater than 0, the negative contribution that reduces the deviation is defined as the effective contribution. i When C is less than 0, a positive contribution that increases the deviation is defined as an effective contribution. Therefore, when C... ij With e i When the product is less than 0, the contribution is marked as a same-direction correction contribution. ij With e i When the product of the two values ​​is greater than 0, the contribution is marked as a reverse conflict contribution. For the same deviation component, the absolute values ​​of all in-direction correction contributions and all reverse conflict contributions are summed to obtain the total in-direction contribution P. i and the total amount of reverse contribution N i and according to N i With P i +N i The system calculates the reverse conflict ratio. When the reverse conflict ratio exceeds a preset reverse conflict ratio, such as 5%, the deviation component is marked as having a reverse cancellation conflict. The system also calculates the algebraic sum of the in-direction contributions. When the absolute value of the algebraic sum of the in-direction contributions exceeds the maximum allowable correction amount for the deviation component within one control cycle, the deviation component is marked as having an in-direction excessive superposition conflict, and the ratio of the actual in-direction contribution to the maximum allowable correction amount is used as the excessive superposition intensity. The control conflict characteristics are constituted by the conflict type, conflict ratio, excessive superposition intensity, involved execution unit number, and corresponding contribution value of each deviation component.

[0038] Subsequently, the initial control requirements are adjusted through coordinated constraints based on the characteristics of control conflicts. Specifically, for deviation components that generate reverse offsetting conflicts, suppression processing is performed to reduce their weight in the control requirements of the corresponding execution unit; for deviation components that generate excessive superposition conflicts in the same direction, gain preservation processing is performed. After one adjustment, the contribution distribution matrix is ​​recalculated using the corrected control requirements. If the reverse conflict ratio of any deviation component is still greater than the preset reverse conflict ratio, or the same-direction contribution still exceeds the maximum allowable correction amount, then conflict identification and constraint adjustment are repeated until the maximum number of iterations is reached. After completing the coordinated constraints, the remaining control gap for each deviation component is calculated. This remaining control gap is the difference between the target correction amount for that deviation component and the sum of the current contributions of all execution units. For each execution unit, its remaining control margin is calculated, and the absolute value of the correction influence coefficient of that execution unit in the same direction as the remaining control gap is multiplied by the remaining control margin to obtain the redistribution capability value. The system normalizes the redistribution capability values ​​corresponding to each deviation component, ensuring the sum of the weights of the participating execution units equals 1. The remaining control gap is then allocated to each execution unit according to the normalized weights, and the allocated increment is added to the constrained control requirements. During redistribution, the system restricts each execution unit from exceeding its rated control range and prioritizes execution units with high efficiency in adjusting the target deviation and minimal collateral impact on other deviations. When an execution unit reaches its control limit, the unallocated control gap is redistributed to the remaining execution units that still have control margin. After redistribution, the control requirements of each execution unit are divided by the corresponding rated control quantity to obtain the normalized control requirements. The system then determines whether the predicted correction amount formed by multiplying the influence mapping matrix and the normalized control requirements makes the absolute values ​​of all deviation components less than their pre-adjustment absolute values. When all deviation components show a decreasing trend, and the predicted increment of any deviation component does not exceed a preset proportion of its original absolute deviation value after crossing zero, the multi-deviation component collaborative convergence condition is satisfied, and the current normalized control requirements are determined as the set of consistent control requirements. Through the above processing, the actual contribution of each execution unit to different deviation components can be clearly identified, the reverse cancellation and excessive superposition between execution units can be suppressed, and the remaining adjustment demand can be redistributed within the control capability range, so that the draft, attitude and motion deviations converge to a stable state synchronously.

[0039] Furthermore, the collaborative constraint processing of the initial control requirement set also includes a time-scale decoupling step: Based on the hydrodynamic response hysteresis parameters and the spectral characteristics of the transient disturbance components, the ballast adjustment unit is designated as a slow-varying control loop, and the propulsion control unit and attitude adjustment unit are designated as fast-varying control loops. Before constructing the mapping relationship between the control actions of each execution unit and the draft deviation, attitude deviation, and motion deviation, the coupling influence weights are decomposed into time scales based on the hydrodynamic response hysteresis parameters. The low-frequency components corresponding to the physical response time constant of the ballast adjustment unit are extracted from the coupling influence weights as slow-varying influence weights. The high-frequency components corresponding to the physical response time constants of the propulsion control unit and attitude adjustment unit are extracted from the coupling influence weights as fast-varying influence weights. The influence mapping relationship of the ballast adjustment unit is constructed based on the slow-varying influence weights, and the influence mapping relationship of the propulsion control unit and attitude adjustment unit is constructed based on the fast-varying influence weights. This ensures that the cooperative constraint adjustment process is executed independently in the slow-varying control loop and the fast-varying control loop, respectively, and the ballast adjustment amount output by the slow-varying control loop is used as the reference setpoint of the fast-varying control loop.

[0040] Optionally, the transient disturbance components within the most recent time period are first read. The mean of each transient disturbance sequence is removed and multiplied by a Hanning window. Then, a Fast Fourier Transform is performed to obtain the power spectra of the average draft disturbance, bow and stern draft disturbance, port and starboard draft disturbance, roll disturbance, pitch disturbance, and heave disturbance. Each power spectrum is divided by its total energy and then summed with weights to form the comprehensive disturbance power spectrum. For example, the weights of draft disturbances, attitude disturbances, and motion disturbances can be set to 0.3, 0.4, and 0.3, respectively. The system accumulates the comprehensive disturbance power spectrum energy from low to high frequency. The frequency corresponding to the cumulative energy reaching the first proportion ηL of the total energy is recorded as the low-frequency energy boundary frequency, and the frequency corresponding to the cumulative energy reaching the second proportion ηU of the total energy is recorded as the effective disturbance upper limit frequency. Subsequently, the pre-calibrated physical response time constants TB of the ballast adjustment unit, TP of the propulsion control unit, and TA of the attitude adjustment unit are read. Each physical response time constant is determined according to the time required for the actual output or corresponding hull state change of the execution unit to reach a preset proportion of final stable change after receiving a step command. Simultaneously, τ is read. c The larger of TB and τc is taken as the slow-varying equivalent response time constant Ts, and the larger of TP, TA, and the lag times of roll, pitch, and heave responses is taken as the fast-varying equivalent response time constant Tf. The system sets the low-frequency decomposition cutoff frequency as fs = min[f ηL ,1 / (2πTs)], where f ηL The low-frequency energy boundary frequency is set, and fs is limited to a preset range z; the fast-change control upper limit frequency is set to fu = min[f ηU , 1 / (2πTf)f max ], where fηU To effectively perturb the upper limit frequency, f max The upper limit threshold for the fast-change control frequency is typically 8Hz. When fu is not greater than fs, it is determined that the effective response bandwidth of the current propulsion control unit and attitude adjustment unit is insufficient, and the use of the fast-change loop to compensate for disturbances in this frequency band is suspended, with only the slow-change control loop remaining operational. For each continuously generated coupling influence weight W... ij (k), according to α s =1 - exp(-2πfsΔt) to calculate the low-frequency recursive coefficients, according to α u =1 - exp(-2πfuΔt) to calculate the upper frequency recursive coefficient, then according to W s,ij (k)=W s,ij (k-1)+α s [W ij (k)-W s,ij [k-1] Extract the low-frequency components of the coupling influence weights, and combine all low-frequency components to form the slowly varying influence weight matrix W. s Meanwhile, according to W u,ij (k)=W u,ij (k-1)+α u [W ij (k)-W u,ij [k-1)] Calculate the weighted components cut off at the upper limit frequency of the fast-change control, and then follow W f,ij (k)=W u,ij (k)-W s,ij (k) Extract the high-frequency components located between the low-frequency decomposition cutoff frequency and the fast-change control upper limit frequency, and combine all high-frequency components to form the fast-change influence weight matrix W. f On the first run, W will... s,ij and W u,ij All are initialized to the current coupling influence weight W. ij This allows the filtering process to start continuously.

[0041] After completing the time-scale decomposition, the bow, stern, port, and starboard ballast adjustment channels are grouped into a ballast execution unit set, and the port and starboard propellers and attitude adjustment mechanisms are grouped into a rapid execution unit set. The ballast influence matrix GB and rapid influence matrix GF are extracted from the pre-calibrated execution unit influence matrix G. GB describes the unit influence of each ballast adjustment channel on the average draft, bow-stern draft difference, port-starboard draft difference, and associated attitude changes. GF ​​describes the unit influence of the propulsion control unit and attitude adjustment unit on roll, pitch, heave, angular velocity, and associated draft changes. The system operates according to HB = (I + W) s GB constructs a slowly varying influence mapping matrix for the ballast regulation unit, according to HF = (I + W) fThe GF constructs a rapidly varying influence mapping matrix for the propulsion control unit and attitude adjustment unit, where I is an identity matrix with the same dimension as the state deviation. This ensures that each influence mapping relationship simultaneously includes the direct control action of the actuators and the indirect control action formed by the transmission of this control action through draft, attitude, and motion coupling. The system uses the same low-pass recursive process as the slow-varying influence weights to process the state deviation, obtaining the slow-varying state deviation e. s and according to e f =ee s Obtain the rapid state deviation e f The slow-change control loop uses HB and the initial ballast control requirement u. B,0 The product approximates the slowly varying state deviation e s To achieve the target, constrained least squares calculations are performed within the limits of ballast pump flow rate, valve opening, and ballast tank capacity to obtain the initial control requirements of the ballast regulating unit. The fast-change control loop does not directly use the original fast-change state deviation; instead, after the slow-change control requirements are determined, the expected change in hull state ΔXB = H caused by ballast regulation is first calculated. B u B And according to XB=X current +ΔXB forms the baseline hull state after ballast adjustment, and then according to eF=X target -XB calculates the remaining state deviation that the fast execution unit needs to compensate, and then uses HF and fast initial control requirements u F,0 The product approximates the remaining state deviation e F The initial control requirements for the propulsion control unit and attitude adjustment unit are obtained, where X current X represents the current stable state vector of the ship. target This represents the target hull stability vector. Through this processing, the expected adjustment results of the ballast adjustment on the mean draft, trim, and list are directly used as the reference setpoint for the fast-change control loop. The fast-change control loop only compensates for the attitude and motion residuals above the reference state and does not reverse the low-frequency draft reference formed by the ballast control.

[0042] The slow-change control loop and the fast-change control loop independently execute control contribution projection, conflict identification, cooperative constraints, and control demand reallocation. Within the slow-change control loop, the contribution of the j-th ballast adjustment channel to the i-th slow-change deviation component is calculated. When the contribution direction is consistent with the corresponding state deviation direction, it is recorded as an effective correction contribution; when the contribution direction is opposite, it is recorded as a reverse cancellation contribution. When the total unidirectional contribution of multiple ballast channels exceeds the maximum allowable correction amount of the deviation component within a slow-change control cycle, it is recorded as unidirectional excessive superposition. The corresponding ballast adjustment demand is then reduced according to the reverse contribution ratio, and the unfinished slow-change control demand is normalized and reallocated according to the remaining ballast capacity and unit adjustment efficiency. Within the fast-change control loop, the contributions of each propulsion control channel and attitude adjustment channel to each fast deviation component are calculated. Reverse cancellation and unidirectional excessive superposition between the propeller and attitude adjustment mechanism are identified in the same way. Conflicting fast control demands are scaled, and the remaining control demands are normalized and reallocated according to the remaining thrust margin, attitude adjustment margin, and unit compensation efficiency. The system updates the slow-change control loop at a first preset frequency, converting the final ballast control requirements into target flow rates and valve openings for each ballast pump. It then updates the fast-change control loop at a second preset frequency, converting the final rapid control requirements into target thrust and thrust direction for the port and starboard thrusters and target control angles for the attitude adjustment mechanism. After each slow-change control update, the reference hull state is synchronously updated and sent to the fast-change control loop. The fast-change control loop maintains the most recent reference hull state between two slow-change update cycles, correcting the rapid control values ​​only based on real-time attitude and motion feedback. Through this process, the coupling effects can be layered according to the actual response speed of the actuators and the frequency distribution of transient disturbances. This allows the ballast adjustment unit to handle low-frequency, continuous draft and static attitude deviations, while the propulsion control unit and attitude adjustment unit handle high-frequency, short-term attitude and motion disturbances. Furthermore, it prevents the fast-change control actions from offsetting the stable draft state formed by ballast adjustment.

[0043] Furthermore, based on the aforementioned control conflict characteristics, the initial control requirements of each execution unit are adjusted through coordinated constraints, including: The control actions of each execution unit on the same deviation component are judged for directional consistency, and the control components that produce reverse cancellation are identified as conflicting components. The conflicting components are suppressed to reduce their proportion weight in the control requirements of the corresponding execution unit. The control components that form a unidirectional superposition effect on the same deviation component are subjected to gain preservation processing. After the conflicting components are suppressed and the unidirectional components are enhanced, the control requirements of each execution unit are recombined to complete the coordinated constraint adjustment.

[0044] Optionally, the state deviation corresponding to the current loop, the influence mapping relationship of each execution unit, and the initial control requirements of each execution unit are read. The unit influence intensity of each execution unit on a certain deviation component is multiplied by the initial control requirements of that execution unit to obtain the control contribution of that execution unit to the deviation component. This yields the contribution direction and magnitude of each execution unit to the average draft deviation, bow-stern draft difference deviation, port-side draft difference deviation, roll angle deviation, pitch angle deviation, heave motion deviation, and angular velocity deviation. Since the state deviation is determined by subtracting the current state value from the target state value, when the direction of a control contribution is the same as the direction of the corresponding state deviation, it indicates that the control contribution can push the current state closer to the target state, and it is marked as a unidirectional control component. When the direction of a control contribution is opposite to the direction of the corresponding state deviation, it indicates that the control contribution will cause the current state to deviate from the target state, or cancel the effective control effect already formed by other execution units, and it is marked as a conflicting component. For the same deviation component, the absolute values ​​of all unidirectional control components and all conflicting components are summed separately. The sum of the conflicting components is then divided by the sum of the two types of control components to obtain the reverse conflict ratio of the deviation component. When the reverse conflict ratio is not greater than the preset reverse conflict ratio, the current reverse action is considered small, and no special suppression is performed. Instead, the corresponding conflicting component, execution unit number, contribution size, and reverse conflict ratio are recorded as control conflict characteristics. Subsequently, the degree of suppression is determined according to the proportion of each conflicting component in the total reverse control action. That is, the weight of the conflicting component in the control requirements of the corresponding execution unit is reduced, so that the adjusted conflicting component is lower than the unadjusted conflicting component, but at least 20% of the unadjusted control requirements are retained. When the same execution unit has a reverse action on the current deviation component, but has a necessary unidirectional action on other deviation components with higher priority such as roll angle, pitch angle, or average draft, the minimum retention ratio of the execution unit is increased to 30% to avoid destroying the effective control of other deviation components while eliminating local conflicts. For the same-direction control component, its original control gain remains unchanged, so that the same-direction control component does not decrease synchronously with the suppression of conflicting components.

[0045] After the conflict component is suppressed, the total effect of the same-direction control component is compared with the correction amount required for the corresponding deviation component. When the same-direction control effect is insufficient and the relevant execution unit still has control margin, the same-direction control component is appropriately increased according to the proportion of the control gap to the original same-direction control effect, but the increased control gain does not exceed the preset multiple of the original control gain. When the same-direction control effect has reached the required correction amount, the original control gain is maintained. When the sum of the same-direction control effects of multiple execution units exceeds the maximum allowable correction amount of the deviation component in one control cycle, the relevant same-direction control component is synchronously reduced according to the proportion between the maximum allowable correction amount and the current total same-direction contribution to prevent over-adjustment. After conflict component suppression and unidirectional component gain processing, control requirements are recombined according to the execution unit number. For each execution unit, the adjusted control contribution of that unit on each deviation component is read. Each adjusted control contribution is divided by the unit's unit influence intensity on the corresponding deviation component to obtain the candidate control requirement corresponding to that deviation component. Then, combination weights are set according to the current deviation magnitude, allowable deviation range, and control priority of each deviation component. The closer the deviation is to or exceeds the allowable range, the greater its combination weight. The weights of average draft deviation, roll deviation, and pitch deviation are higher than those of general angular velocity deviation. The system performs a weighted average of the candidate control requirements of the same execution unit according to each combination weight, and retains a preset proportion of the original initial control requirements as a continuity maintenance quantity to obtain the recombined control requirements of that execution unit. The recombined ballast adjustment requirements must not exceed the allowable operating range of the ballast pumps and ballast tanks, the propulsion control requirements must not exceed the rated thrust range of the propellers, and the attitude adjustment requirements must not exceed the allowable stroke of the attitude adjustment mechanism. After one recombination, the control contribution of each execution unit to each deviation component is recalculated based on the recombined control requirements, and directional consistency is re-evaluated. The collaborative constraint adjustment ends when the reverse conflict ratio of all deviation components is no greater than the preset reverse conflict ratio, and the unidirectional control actions do not exceed the maximum allowable correction amount per cycle for the corresponding deviation component. If significant reverse cancellation or excessive unidirectional superposition still exists, the current recombined control requirements are used as the input for the next adjustment, and conflict identification, conflict component suppression, unidirectional component gain processing, and control requirement recombination are repeated until the maximum number of iterations is reached. Through the above processing, the control requirements of each execution unit constitute the control requirement set after collaborative constraint adjustment, thereby reducing reverse cancellation between different execution units while retaining effective unidirectional control actions and avoiding excessive superposition of multiple execution units on the same deviation component.

[0046] Furthermore, by utilizing the target stability window to configure a stability control strategy and performing unmanned surface vessel (USV) stability adjustments, including: Perform stability adjustment and monitoring of the unmanned surface vessel (USV) and configure monitoring feedback; conduct USV stability verification based on the monitoring feedback; if the USV stability verification fails, generate a USV landing control warning signal and communicate the USV landing control warning signal to the USV waiting to land.

[0047] Preferably, during the coordinated adjustment of the ballast adjustment unit, propulsion control unit, and attitude adjustment unit, hull status information and actuator feedback information are collected at a unified processing frequency, and monitoring feedback is generated within the monitoring feedback window duration. The hull status information includes average draft deviation, bow-stern draft difference, port-ship draft difference, roll angle, pitch angle, roll rate, pitch rate, and heave motion status. The actuator feedback information includes actual ballast output, actual propulsion output, actual attitude adjustment output, and fault status. The system compares the actual output of each actuator with the corresponding control command to obtain the ballast execution tracking deviation, propulsion execution tracking deviation, and attitude execution tracking deviation. When the continuous data loss time exceeds the monitoring data loss tolerance, the time synchronization deviation exceeds the monitoring time synchronization deviation threshold, or an actuator feedback fault occurs, the corresponding monitoring feedback is marked as invalid. Subsequently, based on monitoring feedback, unmanned surface vessel stability verification was performed. The average draft deviation, bow-stern draft difference, port-ship draft difference, roll angle, pitch angle, roll rate, and pitch rate were determined to meet the corresponding allowable thresholds. Based on the platform motion continuity constraint threshold set, it was determined whether the changes in draft, attitude, and motion at adjacent moments were continuous. At the same time, it was determined whether each tracking deviation met the corresponding deviation threshold, and whether the proportion of effective monitoring data reached the effective monitoring data proportion threshold. When the above conditions continuously meet the stable state duration threshold, the unmanned surface vessel (USV) is deemed to have passed the stability verification. When the hull state exceeds the limit for a continuous period of time exceeding the limit, the execution tracking deviation exceeds the limit for a continuous period of time exceeding the execution tracking abnormality threshold, the platform motion continuity constraint is not met, the execution unit malfunctions, the proportion of effective monitoring data is insufficient, or the remaining time of the target stability window is less than the minimum landing completion time of the UAV, the USV is deemed to have failed the stability verification. The target stability window validity flag is removed, a UAV landing control warning signal containing the verification failure reason and landing control command is generated, and the warning signal is sent to the UAV waiting to land to control the UAV to pause descent, exit the approach, or perform a go-around, thereby reducing the landing risk caused by sudden changes in hull attitude or abnormal control execution.

[0048] Furthermore, if the unmanned vessel fails the stability verification for multiple consecutive monitoring cycles, the target stability window will be forcibly unlocked, and the landing capability determination step will be re-executed to construct a new target stability window.

[0049] Optionally, a continuous verification failure counter and a continuous verification failure cycle threshold are set, and the unmanned surface vessel (USV) stability verification results are recorded at the end of each monitoring cycle. When the stability verification of the current monitoring cycle passes, the continuous verification failure counter is reset to zero, and the target stability window remains locked. When the stability verification of the current monitoring cycle fails, the continuous verification failure counter is incremented, and monitoring feedback for the next monitoring cycle is collected. When the continuous verification failure counter reaches the continuous verification failure cycle threshold, it is determined that the current target stability window no longer has recovery value. The locked state of the target stability window is forcibly released, its status is changed from valid to invalid, the output of UAV landing permission signals based on the window is stopped, and the tracking of the target hull status corresponding to the window is stopped. If the UAV to be landed has already entered the landing approach or descent phase, a UAV landing control warning signal is simultaneously sent to the UAV to be landed, instructing it to suspend descent, exit the approach, or perform a go-around. Subsequently, the start and end times, target states, and evaluation results of the original target stability window are cleared, but valid draft, attitude, surface disturbance, and execution unit feedback data from the most recent historical time window are retained. The hull state characterization sequence, hydrodynamic response lag parameters, transient disturbance components, and steady-state response components are updated using the latest monitoring feedback. Then, the hull stability state evolution sequence is regenerated, and landing capability is determined again according to attitude stability constraints, platform motion continuity constraints, the minimum duration threshold of the effective stability window, and the target stability window selection parameters. To avoid reselecting the same time interval that has already failed verification, the original target stability window and candidate stability intervals that overlap with it are excluded, and a new target stability window is determined only from candidate stability intervals whose start time is later than the current time and which meet all stability conditions. When a new target stability window is reconstructed, the continuous verification failure counter is cleared, the new target stability window is locked, and the hull stability control variables are reconfigured according to the new target state. When a new target stability window that meets the requirements is not obtained, the UAV landing permission is kept locked, and stability adjustment monitoring and landing feasibility determination are continued until a new target stability window is constructed, thereby improving the timeliness of target stability window updates and the safety of UAV landing control.

[0050] like Figure 2 As shown, the confusion matrix of the landability determination results is presented, where the horizontal axis represents the predicted result and the vertical axis represents the actual state. The results show that the system has high recognition accuracy for both landable and non-landable states, and the number of correctly determined samples is significantly greater than the number of incorrectly determined samples, indicating that the landability determination method can reliably distinguish whether an unmanned surface vessel meets the landing conditions for an unmanned aerial vehicle.

[0051] In summary, the embodiments of this application have at least the following technical effects: First, multi-source sensor data from the unmanned surface vessel (USV) is read, including draft measurement data, attitude measurement data, and surface disturbance perception data. A unified hull state representation sequence is constructed based on this data. Next, the hull state representation sequence is time-synchronized to construct attitude change and draft change sequences. Time correlation analysis is performed on these sequences to extract the time lag relationship between attitude change and draft change responses, generating hydrodynamic response lag parameters. Then, the hull state representation sequence is decomposed using these lag parameters to establish transient disturbance and steady-state response components. Next, the portion of the steady-state response component that meets preset stability conditions is extracted as the current stable hull state. Time extrapolation is performed using the hydrodynamic response lag parameters combined with the variation characteristics of the transient disturbance component within a historical time window to generate an evolution sequence of the hull stable state. Then, based on the evolution sequence, attitude stability constraints, and platform motion continuity constraints, a landing capability determination is performed to construct a target stability window. Finally, a stability control strategy is configured using the target stability window to perform unmanned surface vessel stability adjustment. This technology solves the technical problem that existing unmanned surface vessels (USVs) cannot accurately detect draft changes in real time and dynamically adjust their attitude under complex sea conditions, resulting in low draft perception accuracy and the inability to make advance stable adjustments for the needs of UAV deployment and recovery operations. It achieves the technical effect of improving the stability and landing capability of USVs in dynamic sea conditions and ensuring operational safety by fusing multi-source sensor data and analyzing the hydrodynamic response lag relationship between attitude changes and draft changes.

[0052] Example 2, based on the same inventive concept as the multi-source sensor data fusion method for real-time draft detection and adjustment of unmanned vessels in the aforementioned examples, such as... Figure 3 As shown, this application provides a real-time draft detection and adjustment system for unmanned vessels based on multi-source sensor data fusion, wherein the system includes: Data reading unit 11: Reads multi-source sensor data from the unmanned surface vessel, including draft measurement data, attitude measurement data, and surface disturbance perception data, and constructs a unified hull state representation sequence based on the multi-source sensor data; Time correlation analysis unit 12: Performs time synchronization processing on the hull state representation sequence, constructs an attitude change sequence and a draft change sequence, performs time correlation analysis on the attitude change sequence and the draft change sequence, extracts the time lag relationship between the attitude change response and the draft change response, and generates hydrodynamic response lag parameters; Sequence decomposition unit 13: Decomposes the hull state using the hydrodynamic response lag parameters. The system decomposes the state representation sequence to establish transient disturbance components and steady-state response components; Time extrapolation unit 14: extracts the part of the steady-state response component that meets the preset stability conditions as the current stable state of the hull, and uses the hydrodynamic response lag parameter combined with the change characteristics of the transient disturbance component in the historical time window to perform time extrapolation to generate the evolution sequence of the hull stable state; Landability determination unit 15: performs landing capability determination based on the evolution sequence, attitude stability constraints, and platform motion continuity constraints, and constructs a target stability window; Stability adjustment unit 16: configures the stability control strategy using the target stability window and performs unmanned surface vessel stability adjustment.

[0053] Furthermore, the data reading unit 11 is used to perform the following method: the acquisition of the surface disturbance sensing data includes at least one of the following: a: by using a millimeter-wave radar, lidar, or monocular / binocular vision sensor installed on the hull, the wave height sequence or wave image sequence around the ship is collected, and the wave height, wave period, and wave direction parameters are extracted as surface disturbance sensing data; b: by using the shipborne inertial wave sensing module, the wave statistical parameters at the location of the hull are obtained based on the vertical acceleration and angular velocity sequence inversion of the six-degree-of-freedom IMU as surface disturbance sensing data; the surface disturbance sensing data is aligned with the draft measurement data and attitude measurement data on the time axis to construct the hull state characterization sequence.

[0054] Furthermore, the stability adjustment unit 16 is used to perform the following method: read the state deviation between the target stability window and the current hull stability state; configure the hull stability control quantity according to the state deviation, and send the hull stability control quantity to the ballast adjustment unit, the propulsion control unit and the attitude adjustment unit to perform coordinated adjustment.

[0055] Furthermore, the stability adjustment unit 16 is used to perform the following methods: constructing a coupling relationship representation between draft deviation, attitude deviation, and motion deviation based on the state deviation, wherein the coupling relationship representation is used to describe the mutual influence direction and influence intensity between each deviation component; performing time-series correction on the coupling relationship representation in conjunction with the hydrodynamic response hysteresis parameter to obtain the coupling influence weight of each deviation component at different time scales; mapping the state deviation to an initial control demand set of multiple execution units based on the coupling influence weight, wherein the control demand corresponding to any execution unit simultaneously includes the adjustment effect on multiple deviation components; performing collaborative constraint processing on the initial control demand set to suppress control conflicts caused by coupling effects between different execution units, and redistributing the control contribution of each execution unit to generate a consistent control demand set; constructing hull stability control quantities according to the consistent control demand set, and distributing the hull stability control quantities to the ballast adjustment unit, propulsion control unit, and attitude adjustment unit respectively for collaborative adjustment.

[0056] Furthermore, the stabilization adjustment unit 16 is used to perform the following methods: constructing a mapping relationship between the control actions of each execution unit and the draft deviation, attitude deviation, and motion deviation based on the coupling influence weight; projecting the initial control requirements of each execution unit onto the influence mapping relationship to obtain the contribution distribution of each execution unit to each deviation component; identifying the unidirectional superposition and reverse cancellation effects between different execution units on the same deviation component based on the contribution distribution of each execution unit, and determining the control conflict characteristics; adjusting the initial control requirements of each execution unit according to the control conflict characteristics; and normalizing and redistributing the control requirements after constraint adjustment to generate a set of consistent control requirements that satisfy the conditions for the coordinated convergence of multiple deviation components.

[0057] Furthermore, the stabilization adjustment unit 16 is used to perform the following method: based on the hydrodynamic response hysteresis parameters and the spectral characteristics of the transient disturbance components, the ballast adjustment unit is set as a slow-varying control loop, and the propulsion control unit and attitude adjustment unit are set as fast-varying control loops; before constructing the influence mapping relationship of the control actions of each execution unit on draft deviation, attitude deviation, and motion deviation, the coupling influence weights are decomposed on a time scale according to the hydrodynamic response hysteresis parameters; the low-frequency component corresponding to the physical response time constant of the ballast adjustment unit in the coupling influence weights is extracted as the slow-varying influence weight; the high-frequency component corresponding to the physical response time constant of the propulsion control unit and attitude adjustment unit in the coupling influence weights is extracted as the fast-varying influence weight; the influence mapping relationship of the ballast adjustment unit is constructed based on the slow-varying influence weights, and the influence mapping relationship of the propulsion control unit and attitude adjustment unit is constructed based on the fast-varying influence weights; so that the cooperative constraint adjustment process is executed independently in the slow-varying control loop and the fast-varying control loop respectively, and the ballast adjustment amount output by the slow-varying control loop is used as the reference setpoint of the fast-varying control loop.

[0058] Furthermore, the stabilization adjustment unit 16 is used to perform the following method: determine the directional consistency of the control actions of each execution unit on the same deviation component, and identify the control components that produce reverse cancellation effects as conflicting components; perform suppression processing on the conflicting components to reduce their proportion weight in the control requirements of the corresponding execution unit; perform gain maintenance processing on the control components that form unidirectional superposition effects on the same deviation component; after completing the suppression of conflicting components and the enhancement of unidirectional components, recombin the control requirements of each execution unit to complete the coordinated constraint adjustment.

[0059] Furthermore, the stabilization adjustment unit 16 is used to perform the following method: determine the directional consistency of the control actions of each execution unit on the same deviation component, and identify the control components that produce reverse cancellation effects as conflicting components; perform suppression processing on the conflicting components to reduce their proportion weight in the control requirements of the corresponding execution unit; perform gain maintenance processing on the control components that form unidirectional superposition effects on the same deviation component; after completing the suppression of conflicting components and the enhancement of unidirectional components, recombin the control requirements of each execution unit to complete the coordinated constraint adjustment.

[0060] Furthermore, the stabilization adjustment unit 16 is used to perform the following method: If the unmanned vessel fails the stability verification for multiple consecutive monitoring cycles, the target stability window will be forcibly unlocked, and the landing capability determination step will be re-executed to construct a new target stability window.

[0061] 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 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 real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion, characterized in that, The method includes: Read multi-source sensor data from the unmanned vessel, including draft measurement data, attitude measurement data, and surface disturbance perception data, and construct a unified hull state representation sequence based on the multi-source sensor data; The ship's state characterization sequence is time-synchronized to construct attitude change sequence and draft change sequence. Time correlation analysis is performed on attitude change sequence and draft change sequence to extract the time lag relationship between attitude change response and draft change response and generate hydrodynamic response lag parameters. The hull state characterization sequence is decomposed using hydrodynamic response hysteresis parameters to establish transient disturbance components and steady-state response components. The portion of the steady-state response component that meets the preset stability condition is extracted as the current stable state of the hull. The evolution sequence of the stable state of the hull is generated by extrapolating the time using the hydrodynamic response lag parameter combined with the variation characteristics of the transient disturbance component in the historical time window. Based on the evolution sequence, attitude stability constraints, and platform motion continuity constraints, a landing capability determination is performed, and a target stability window is constructed. The target stability window is used to configure the stability control strategy and execute the stability adjustment of the unmanned vessel.

2. The method for real-time draft detection and adjustment of unmanned vessels based on multi-source sensor data fusion as described in claim 1, characterized in that, Utilizing the target stability window to configure stability control strategies, unmanned surface vessel (USV) stability adjustments are executed, including: Read the state deviation between the target stability window and the current hull stability state; Based on the aforementioned state deviation, configure the hull stability control quantity and send the hull stability control quantity to the ballast adjustment unit, propulsion control unit, and attitude adjustment unit to perform coordinated adjustment.

3. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 2, characterized in that, Configure hull stability control parameters based on the aforementioned state deviation, including: Based on the state deviation, a coupling relationship characterization is constructed between draft deviation, attitude deviation, and motion deviation. The coupling relationship characterization is used to describe the mutual influence direction and influence intensity between each deviation component. By combining the hydrodynamic response lag parameters, the coupling relationship characterization is time-series corrected to obtain the coupling influence weights of each deviation component at different time scales. Based on the coupling influence weight, the state deviation is mapped to a set of initial control requirements for multiple execution units, wherein the control requirement corresponding to any execution unit simultaneously includes the adjustment effect on multiple deviation components. The initial set of control requirements is subjected to collaborative constraint processing to suppress control conflicts caused by coupling effects between different execution units, and the control contributions of each execution unit are redistributed to generate a consistent set of control requirements. Based on the set of consistent control requirements, the hull stability control quantities are constructed, and the hull stability control quantities are respectively sent to the ballast adjustment unit, the propulsion control unit and the attitude adjustment unit to perform coordinated adjustment.

4. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 3, characterized in that, The initial set of control requirements is subjected to coordinated constraint processing, including: Based on the aforementioned coupling influence weights, a mapping relationship is constructed between the influence of the control actions of each execution unit on draft deviation, attitude deviation, and motion deviation. The initial control requirements of each execution unit are projected onto the influence mapping relationship to obtain the contribution distribution of each execution unit to each deviation component. Based on the contribution distribution of each execution unit on the same deviation component, identify the synergistic and counteracting effects of different execution units on the deviation component, and determine the characteristics of control conflicts. Based on the aforementioned control conflict characteristics, the initial control requirements of each execution unit are adjusted in a coordinated manner. The control requirements after constraint adjustment are normalized and redistributed to generate a set of consistent control requirements that satisfy the conditions for coordinated convergence of multiple deviation components.

5. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 4, characterized in that, The collaborative constraint processing of the initial control requirement set also includes a time-scale decoupling step: Based on the hydrodynamic response hysteresis parameters and the spectral characteristics of the transient disturbance components, the ballast adjustment unit is set as a slow-change control loop, and the propulsion control unit and attitude adjustment unit are set as fast-change control loops. Before constructing the mapping relationship between the control actions of each execution unit and the draft deviation, attitude deviation and motion deviation, the coupling influence weights are decomposed on a time scale according to the hydrodynamic response hysteresis parameters. The low-frequency component corresponding to the physical response time constant of the ballast regulating unit is extracted from the coupling influence weight and used as the slowly varying influence weight; The high-frequency components corresponding to the physical response time constants of the propulsion control unit and attitude adjustment unit in the coupling influence weights are extracted and used as the fast-changing influence weights. The influence mapping relationship of the ballast adjustment unit is constructed based on the slow-changing influence weight, and the influence mapping relationship of the propulsion control unit and attitude adjustment unit is constructed based on the fast-changing influence weight. This allows the coordinated constraint adjustment process to be executed independently in the slow-change control loop and the fast-change control loop, respectively, with the ballast adjustment amount output by the slow-change control loop serving as the reference setting value for the fast-change control loop.

6. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 4, characterized in that, Based on the aforementioned control conflict characteristics, the initial control requirements of each execution unit are coordinated and constrained, including: The direction consistency of the control action of each execution unit on the same deviation component is judged, and the control component that produces the reverse cancellation effect is identified as the conflict component. Suppression processing is performed on the conflicting components to reduce their weight in the control requirements of the corresponding execution unit. Gain preservation processing is performed on control components that have a unidirectional superposition effect on the same deviation component; After suppressing conflict components and enhancing unidirectional components, the control requirements of each execution unit are recombined to complete the adjustment of collaborative constraints.

7. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 1, characterized in that, The water surface disturbance sensing data is acquired through at least one of the following methods: a: By using millimeter-wave radar, lidar or monocular / binocular vision sensors installed on the hull, collect wave height sequence or wave image sequence around the ship, and extract wave height, wave period and wave direction parameters as water surface disturbance sensing data. b: By using the shipborne inertial wave sensing module, the wave statistical parameters of the ship's location are obtained as water surface disturbance sensing data based on the vertical acceleration and angular velocity sequence inversion of the six-degree-of-freedom IMU. Align the surface disturbance sensing data with the draft measurement data and attitude measurement data on the time axis to construct the hull state representation sequence.

8. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 1, characterized in that, Utilizing the target stability window to configure stability control strategies, unmanned surface vessel (USV) stability adjustments are executed, including: Perform stability monitoring and adjustment of unmanned vessels, and configure monitoring feedback; The unmanned vessel stability verification is performed based on the monitoring feedback. If the unmanned vessel stability verification fails, a drone landing control warning signal is generated and the drone landing control warning signal is communicated to the drone waiting to land.

9. The method for real-time draft detection and adjustment of unmanned surface vessels based on multi-source sensor data fusion as described in claim 8, characterized in that, If the unmanned vessel fails the stability verification for multiple consecutive monitoring cycles, the target stability window will be forcibly unlocked, and the landing capability determination step will be re-executed to construct a new target stability window.

10. A real-time draft detection and adjustment system for unmanned vessels based on multi-source sensor data fusion, characterized in that: The system is used to implement the multi-source sensor data fusion method for real-time draft detection and adjustment of an unmanned vessel according to any one of claims 1-9, the system comprising: Data reading unit: reads multi-source sensor data of the unmanned vessel, including draft measurement data, attitude measurement data and surface disturbance perception data, and constructs a unified hull state representation sequence based on the multi-source sensor data; Time correlation analysis unit: Performs time synchronization processing on the hull state characterization sequence, constructs attitude change sequence and draft change sequence, performs time correlation analysis on attitude change sequence and draft change sequence, extracts the time lag relationship between attitude change response and draft change response, and generates hydrodynamic response lag parameters. Sequence decomposition unit: The sequence characterization of the hull state is decomposed using hydrodynamic response lag parameters to establish transient disturbance components and steady-state response components; Time extrapolation unit: Extracts the portion of the steady-state response component that meets the preset stability condition as the current stable state of the hull, and uses the hydrodynamic response lag parameter combined with the variation characteristics of the transient disturbance component in the historical time window to perform time extrapolation, generating the evolution sequence of the stable state of the hull. Landability determination unit: performs landingability determination based on the evolution sequence, attitude stability constraints, and platform motion continuity constraints, and constructs a target stability window; Stability Adjustment Unit: Utilizes the target stability window to configure the stability control strategy and executes the stability adjustment of the unmanned vessel.

Images