An end-cloud cooperative inland river multi-ship virtual formation and distance keeping method

Abstract

The present application relates to the field of inland river intelligent shipping cooperative control, and discloses an end-cloud cooperative inland river multi-ship virtual formation and distance keeping method, which comprises the following steps: calculating the residual sequence between the following ship longitudinal acceleration sequence in the front and rear ship longitudinal following sample sequence, the single ship longitudinal acceleration sequence and the reference longitudinal hydrodynamic action item sequence as the longitudinal interaction residual sequence; generating the longitudinal hydrodynamic action item correction coefficient in the cloud to construct the longitudinal hydrodynamic action item online correction coefficient table; embedding the longitudinal hydrodynamic action item online correction coefficient table generated in the cloud into the following ship prediction model preset in the ship end of the following ship; solving the distance keeping quadratic programming problem with propulsion constraint to output the following ship propulsion control instruction for stably keeping the longitudinal distance between ships. The present application realizes the longitudinal distance stable keeping in the inland river multi-ship virtual formation through the prediction generation of the following ship propulsion control instruction based on the end-cloud cooperative ship-to-ship hydrodynamic interaction modeling and online correction.

Description

Technical Field

[0001] This invention relates to the fields of intelligent inland waterway shipping, collaborative control of multi-ship virtual formations, and predictive control of ship models, and particularly to a method for maintaining spacing in inland waterway multi-ship virtual formations through end-to-cloud collaboration. Background Technology

[0002] As inland waterway transportation develops towards intelligence, efficiency, and automation, the traditional single-vessel autonomous navigation mode can no longer meet the organizational needs of high-density, high-efficiency waterway transportation. Multi-vessel virtual formation technology enables multiple vessels to maintain a stable formation without physical connection through information sharing and collaborative control among vessels. This technology can be used in inland waterway scenarios such as lock passage, narrow channels, and port area scheduling.

[0003] Virtual formation technology enables multiple ships to maintain a preset formation and relative distance without physical connection by establishing information interaction relationships between them. This reduces the impact of ship obstruction and improves waterway utilization and transportation organization efficiency. Among these, maintaining longitudinal spacing is a key aspect of multi-ship virtual formation control, requiring the following ships to adjust their propulsion control strategies in real time based on the motion status of the preceding ship, the hydrodynamic interaction between ships, and their own dynamic response characteristics.

[0004] Model predictive control (MDC) methods can handle ship dynamic constraints, environmental disturbances, and future state prediction problems, and are therefore often used for virtual formation spacing control. However, existing MDC methods based on fixed-parameter hydrodynamic models typically rely on offline identified ship dynamic parameters and ship-to-ship interaction models. Due to the influence of ship type differences, current disturbances, shallow water effects, and hydrodynamic coupling caused by close-range navigation in inland waterways, the longitudinal motion of ships exhibits significant nonlinear and time-varying characteristics. Traditional fixed-model-based spacing control methods struggle to meet the high-precision formation control requirements under complex conditions. When there are differences in ship type, changes in load capacity, or changes in actual navigation conditions within the formation, deviations will occur between the actual hydrodynamic coupling effects and the model predictions.

[0005] Furthermore, for close-following convoys composed of inland cargo ships of different types or with different load conditions, the coefficients of the fixed ship-to-ship hydrodynamic interaction model are prone to mismatch with the actual ships, resulting in longitudinal relative speed prediction errors, and further causing continuous deviations or short-term underruns in the expected spacing. If this problem is circumvented simply by adjusting the target spacing or changing the formation structure, it will reduce the channel utilization efficiency and formation stability of the virtual formation.

[0006] To address this issue, this invention proposes a cloud-edge collaborative method for virtual formation and spacing maintenance of multiple vessels in inland waterways. By adaptively correcting the longitudinal action term of the ship-to-ship hydrodynamic interaction model during close following of heterogeneous inland waterway vessels, the following vessels can stably maintain the longitudinal distance between the vessels without changing the formation structure and target spacing settings. Summary of the Invention

[0007] This invention proposes a cloud-edge collaborative method for virtual formation and spacing maintenance of multiple vessels in inland waterways. Step S1 involves unifying the time reference, transforming spatial coordinates, and calculating longitudinal projections of multi-source vessel state data to form a sequence of longitudinal following samples containing longitudinal spacing and relative longitudinal velocity, achieving a unified representation of the motion states of the vessels. Step S2 involves jointly calculating the spatiotemporally registered longitudinal following samples with a single-vessel dynamic response model and a benchmark longitudinal hydrodynamic action model, extracting the interaction residuals between the actual longitudinal acceleration of the following vessel and the theoretical single-vessel response and benchmark hydrodynamic action, thus quantifying the hydrodynamic influence characteristics between vessels and solving the problem that traditional vessel motion models cannot accurately describe the hydrodynamic interaction during close-range navigation between vessels. Step S3 introduces a continuous spacing variation... The system employs constraints and residual fitting to perform online correction of the baseline longitudinal hydrodynamic coefficient table, enabling adaptive updates of the baseline longitudinal hydrodynamic coefficients. Step S4 integrates the online correction coefficient table of longitudinal hydrodynamic action terms generated in the cloud with the pre-set prediction model of the following ship at the ship's end. The correction coefficients of the longitudinal hydrodynamic action terms obtained online are introduced into the ship motion state prediction process, improving the following ship prediction model's ability to match the actual formation navigation state and solving the problem of insufficient prediction accuracy caused by the ship-end prediction model's inability to reflect real-time changes in hydrodynamics between ships. Step S5 recursively extrapolates the future ship motion state of the following ship and, in conjunction with propulsion constraints, constructs a quadratic programming optimization problem to maintain the distance, solving for propulsion control commands that meet the target distance requirements, enabling the following ship to actively respond to changes in the motion of the preceding ship.

[0008] To achieve the above objectives, the present invention provides a method for virtual formation and spacing maintenance of multiple vessels in inland waterways through end-to-cloud collaboration, comprising the following steps: S1: Collect the following ship's positioning sampling data output by the RTK positioning receiver at the following ship's end, the leading ship's broadcast sampling data broadcast by the leading ship through the formation data link, the following ship's motion sampling data output by the following ship's inertial measurement unit, and the ship's basic information. Perform time alignment and longitudinal projection calculations on the following ship's positioning sampling data, the leading ship's broadcast sampling data, and the following ship's motion sampling data to form a longitudinal following sample sequence of the fore-and-aft ships that has completed spatiotemporal registration. S2: Based on the longitudinal following sample sequences of the fore and aft ships that have completed spatiotemporal registration, generate a single ship longitudinal acceleration sequence and a reference longitudinal hydrodynamic action term sequence. Calculate the residual sequence between the longitudinal acceleration sequence of the following ship and the longitudinal acceleration sequence of the single ship and the reference longitudinal hydrodynamic action term sequence in the longitudinal following sample sequences of the fore and aft ships. Use this as the longitudinal interactive residual sequence. Concatenate the longitudinal interactive residual sequence to the longitudinal following sample sequences of the fore and aft ships to obtain the longitudinal following sample sequences of the fore and aft ships that have completed the longitudinal interactive residual calibration. S3: Based on the longitudinal interactive residual calibration of the fore and aft ship following sample sequences, the longitudinal hydrodynamic action term correction coefficients are generated in the cloud, and an online correction coefficient table for the longitudinal hydrodynamic action term is constructed. S4: Embed the online correction coefficient table of longitudinal hydrodynamic action terms generated in the cloud into the pre-set prediction model of the following ship at the end of the following ship to form a prediction model of the following ship that completes the online correction of longitudinal hydrodynamic action terms. S5: Based on the state recursion of the follower prediction model with online correction of longitudinal hydrodynamic terms, solve the quadratic programming problem of spacing maintenance with propulsion constraints, and output the follower propulsion control command for maintaining the longitudinal distance between ships stably.

[0009] As a further improvement of the present invention: Furthermore, in step S1, the following ship positioning sampling data includes the timestamp, longitude, latitude, and heading angle data collected by the following ship's RTK positioning receiver; the leading ship broadcast sampling data includes the timestamp, longitude, latitude, heading angle data, and longitudinal velocity of the leading ship broadcast through the formation data link; the following ship motion sampling data includes the longitudinal acceleration of the following ship collected by the following ship's inertial measurement unit, the propulsion speed of the following ship output by the propulsion controller, and the rudder angle of the following ship; and the ship's basic information includes the length of the leading ship, the length of the following ship, the draft of the following ship, the longitudinal distance between the target and the sampling period.

[0010] Furthermore, step S1 involves performing time alignment and longitudinal projection calculations on the following vessel's positioning sampling data, the preceding vessel's broadcast sampling data, and the following vessel's motion sampling data. S11: Perform time base conversion and unified sampling period processing on the timestamps in the following ship positioning sampling data, the preceding ship broadcast sampling data, and the following ship motion sampling data to establish a unified sampling time series, and perform time alignment on various types of sampling data based on the unified sampling time series to obtain a synchronous sampling data series. S12: Convert the latitude and longitude in the synchronous sampling data sequence to planar position coordinates under a unified coordinate reference to form a spatial position sequence of the fore and aft ships that includes sampling time, planar position coordinates of the preceding ship, and planar position coordinates of the following ship. S13: Based on the spatial position sequence of the fore and aft ships and the heading angle data of the following ship at different sampling times, calculate the longitudinal distance between the fore and aft ships at different sampling times, and combine the changes in the longitudinal distance between the fore and aft ships at adjacent sampling times to calculate the longitudinal relative speed between the fore and aft ships at different sampling times. S14: Read the longitudinal distance between the fore and aft ships, the longitudinal relative velocity, the longitudinal acceleration of the following ship, the propulsion speed of the following ship, and the rudder angle of the following ship at different sampling times, and sort the read data according to the order of occurrence of the sampling time to obtain a sorted sequence. The sorted sequence is then concatenated with the length of the fore ship, the length of the following ship, the draft of the following ship, and the longitudinal distance between the target ships to obtain the longitudinal following sample sequence of the fore and aft ships for completing the spatiotemporal registration.

[0011] Specifically, the calculation formulas for the longitudinal distance between the fore and aft ships and the longitudinal relative velocity at different sampling times in step S13 are as follows: ; ; in, This represents the longitudinal distance between the preceding and following vessels at the nth sampling time. This represents the longitudinal relative velocity between the preceding and following vessels at the nth sampling time. Indicates the sampling period. This represents the lateral coordinate of the ship's front plane at the nth sampling time. This represents the longitudinal coordinate of the forward ship at the nth sampling time. This represents the lateral coordinate of the following ship in the plane at the nth sampling time. This represents the longitudinal coordinate of the following ship in the plane at the nth sampling time. This represents the heading angle data of the following ship at the nth sampling time, where N represents the total number of sampling times.

[0012] Further, in step S2, based on the longitudinal following sample sequences of the fore and aft ships that have completed spatiotemporal registration, a single-ship longitudinal acceleration sequence and a reference longitudinal hydrodynamic action term sequence are generated, including: S21: Read the longitudinal distance between the fore and aft ships at different sampling times in the longitudinal following sample sequence of the fore and aft ships, and normalize the longitudinal distance between the fore and aft ships based on the length of the fore ship and the length of the following ship to obtain the normalized longitudinal distance. S22: Receive the reference longitudinal hydrodynamic coefficient table sent from the cloud, perform linear interpolation on the reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficient table based on the normalized longitudinal spacing, obtain the reference longitudinal hydrodynamic action item corresponding to the normalized longitudinal spacing, sort the reference longitudinal hydrodynamic action item based on the sequence order of the normalized longitudinal spacing in the longitudinal following sample sequence of the fore and aft ships, and obtain the reference longitudinal hydrodynamic action item sequence. S23: Read the propulsion speed and rudder angle of the following ship at different sampling times in the longitudinal following sample sequence of the preceding and following ships. Use the read content and the draft value of the following ship as the input values ​​of the single ship longitudinal acceleration calibration curve preset at the end of the following ship. Extract the single ship longitudinal acceleration at the sampling time associated with the read content from the single ship longitudinal acceleration calibration curve. Sort the single ship longitudinal acceleration according to the order of occurrence of the sampling time to obtain the single ship longitudinal acceleration sequence.

[0013] Furthermore, step S22, which involves linearly interpolating the reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficient table based on the normalized longitudinal spacing, further includes: The reference longitudinal hydrodynamic coefficient table consists of several fixed longitudinal spacing nodes with equal spacing and the reference longitudinal hydrodynamic coefficients corresponding to the fixed longitudinal spacing nodes. Two adjacent fixed longitudinal spacing nodes are used as a set of longitudinal spacing grids, and two adjacent fixed longitudinal spacing nodes are used as the left grid spacing point and the right grid spacing point of the longitudinal spacing grid, respectively, where the right grid spacing point is greater than the right grid spacing point. If the normalized longitudinal spacing lies between the left and right grid spacing points in any longitudinal spacing grid, then based on the reference longitudinal hydrodynamic coefficients corresponding to the left and right grid spacing points of that longitudinal spacing grid, a reference longitudinal hydrodynamic term corresponding to the normalized longitudinal spacing is generated, wherein the formula for generating the reference longitudinal hydrodynamic term is: ; in, Indicates normalized vertical spacing The corresponding baseline longitudinal hydrodynamic term, Indicates normalized vertical spacing The left grid spacing point of the vertical grid in which it is located. Indicates normalized vertical spacing The right grid spacing point of the vertical grid in which it is located. Indicates the left grid spacing point The corresponding reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficients table. Indicates the right grid spacing point The corresponding reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficients table. Indicates normalized vertical spacing Corresponding left grid spacing point Spacing coefficient, Indicates normalized vertical spacing Corresponding right grid spacing points Spacing coefficient.

[0014] Further, in step S3, correction coefficients for the longitudinal hydrodynamic action term are generated, and an online correction coefficient table for the longitudinal hydrodynamic action term is constructed, including: S31: Calculate the difference between adjacent normalized longitudinal distances in the preceding and following ship following sample sequences after longitudinal interactive residual calibration is completed. Sort the data at different sampling time points in the preceding and following ship following sample sequences after longitudinal interactive residual calibration is completed in ascending order of the difference to obtain a rearranged result sequence. Based on the difference sequence formed by the difference between adjacent normalized longitudinal distances in ascending order, construct the difference constraint matrix of the longitudinal hydrodynamic action term decaying with distance. S32: Extract the longitudinal interactive residual sequence from the rearranged result sequence as the observation, and use the spacing coefficients of the left grid spacing point and the right grid spacing point corresponding to the normalized longitudinal spacing as the regressors. Solve the weighted least squares equation containing the zero-increment anchoring term and the difference constraint matrix to obtain the longitudinal hydrodynamic action term correction coefficient vector. The longitudinal hydrodynamic action term correction coefficient vector consists of the longitudinal hydrodynamic action term correction coefficients corresponding to the fixed longitudinal spacing nodes after ascending sorting in the benchmark longitudinal hydrodynamic coefficient table. S33: Write the longitudinal hydrodynamic action item correction coefficients in the longitudinal hydrodynamic action item correction coefficient vector into the fixed longitudinal spacing nodes of the standard longitudinal hydrodynamic coefficient table after ascending order, as the online correction coefficient table for the longitudinal hydrodynamic action item.

[0015] Furthermore, in step S4, the online correction coefficient table for the longitudinal hydrodynamic action term generated in the cloud is embedded into the pre-set prediction model of the following vessel at the end of the following vessel, including: S41: Add the online correction coefficient table of longitudinal hydrodynamic action items generated in the cloud to the baseline longitudinal hydrodynamic coefficient table cached at the end of the following ship item by item to calculate the corrected longitudinal hydrodynamic coefficient table; S42: Based on the modified longitudinal hydrodynamic coefficient table, regenerate the modified longitudinal hydrodynamic action terms and the modified longitudinal hydrodynamic action term sequence for different sampling times according to the linear interpolation method described in step S22, and write them into the pre-set follower prediction model at the end of the follower ship to generate the follower prediction model with completed online correction of the longitudinal hydrodynamic action terms.

[0016] Furthermore, in step S5, the following ship prediction model with online correction of the longitudinal hydrodynamic action term is used as the basis for state recursion to solve the spacing-preserving quadratic programming problem with propulsion constraints, including: S51: Using the prediction model of the following ship with online correction of the longitudinal hydrodynamic action term as the state recursive equation, the predicted longitudinal distance between the preceding and following ships and the predicted longitudinal relative speed of the following ship in the prediction time domain are generated. The propulsion speed in the prediction time domain is used as the parameter to be solved. According to the square of the deviation between the predicted longitudinal distance between the preceding and following ships and the target longitudinal distance, the square of the longitudinal relative speed and the square of the change in propulsion speed in the prediction time domain, a quadratic programming objective function for maintaining the distance is constructed. S52: Apply propulsion speed limit and propulsion speed change rate limit to the quadratic programming objective function that maintains spacing, and solve to obtain the candidate sequence of propulsion speed in the predicted time domain; S53: Extract the first propulsion speed from the candidate sequence of predicted time-domain propulsion speeds, and generate a follower propulsion control command based on the extracted propulsion speed to maintain the longitudinal distance between ships, so that the follower ships can maintain the target longitudinal distance under the condition that the longitudinal hydrodynamic action term has been corrected online.

[0017] Compared with existing technologies, this invention proposes a method for virtual formation and spacing maintenance of multiple vessels in inland waterways through end-to-cloud collaboration. This technology has the following beneficial effects: First, this invention achieves accurate correlation of multi-source ship perception data with different sources and sampling frequencies at the same time scale by performing time base conversion, sampling period unification, and time sequence synchronization processing on the positioning sampling data of the following ship, the broadcast sampling data of the preceding ship, and the motion sampling data of the following ship. This avoids the problem of misalignment of ship status information caused by communication delays, equipment clock deviations, and asynchronous sampling, and improves the accuracy of the fusion of the motion status of the preceding and following ships. Specifically, this invention converts the latitude and longitude data of the preceding and following ships to a unified plane coordinate system and combines the heading angle of the following ship to perform longitudinal projection on the spatial position difference between the preceding and following ships, converting the two-dimensional spatial position relationship into a longitudinal distance relationship along the ship's sailing direction. This allows the calculated longitudinal distance between the preceding and following ships to effectively reflect the actual following distance during the ship formation process, reducing the impact of curved channels and changes in heading on the distance measurement results. Furthermore, by using the longitudinal distance change between adjacent sampling times to calculate the longitudinal relative velocity, a continuous characterization of the dynamic motion relationship between the preceding and following ships is achieved. Finally, by fusing longitudinal spacing, longitudinal relative speed, follower ship dynamic response parameters, and ship basic parameters in chronological order, a spatiotemporally registered longitudinal following sample sequence of fore and aft ships was constructed, improving the reliability of virtual multi-ship formation and spacing maintenance.

[0018] Meanwhile, this invention normalizes the longitudinal spacing between the fore and aft ships and generates a baseline longitudinal hydrodynamic action term by combining it with a baseline longitudinal hydrodynamic coefficient table. This achieves a unified representation of the hydrodynamic action between ships in formation scenarios of different ship types and scales, avoiding the problem of the model being difficult to apply across ship types due to directly using the actual longitudinal spacing, and improving the universality and portability of longitudinal hydrodynamic modeling. Furthermore, this invention uses a longitudinal spacing grid and linear interpolation to calculate the baseline longitudinal hydrodynamic action term, ensuring that any normalized longitudinal spacing can yield continuous and smooth baseline hydrodynamic estimates. This avoids parameter jumps caused by direct table lookups, improving the continuity and stability of hydrodynamic action estimation. Further, by generating a single-ship longitudinal acceleration sequence based on the propulsion speed, rudder angle, and draft of the following ship, the dynamic response generated by the ship's own propulsion maneuvering is decoupled from the longitudinal hydrodynamic action between ships. The longitudinal interaction residual sequence is obtained by subtracting the single-ship longitudinal acceleration and the baseline longitudinal hydrodynamic action term from the actual longitudinal acceleration. This residual sequence can more accurately reflect the unmodeled dynamic interaction effects of complex wake coupling, flow field evolution, and environmental disturbances. Attached Figure Description

[0019] Figure 1 A flowchart illustrating a method for virtual formation and spacing maintenance of multiple vessels in inland waterways through end-to-cloud collaboration, as provided in an embodiment of the present invention. Figure 2 A comparison diagram of the technical effects of the virtual formation and spacing maintenance method for multiple vessels on inland waterways provided in an embodiment of the present invention; Figure 3 This is a comparison diagram showing the fitting effect between the predicted trajectory of the following ship and the actual trajectory output by the ship-following prediction model provided in an embodiment of the present invention. Detailed Implementation

[0020] The realization of the objectives, functional characteristics, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] This invention provides a method for virtual formation and spacing maintenance of multiple vessels on inland waterways using an end-to-cloud collaborative approach. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this invention: a server, a terminal, etc. In other words, this method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0022] Reference Figure 1Embodiment 1 of the present invention is as follows: A method for virtual formation and spacing maintenance of multiple vessels in inland waterways using edge-cloud collaboration, the method comprising: S1: Collect the following ship's positioning sampling data output by the RTK positioning receiver on the following ship, the leading ship's broadcast sampling data broadcast by the leading ship through the formation data link, the following ship's motion sampling data output by the following ship's inertial measurement unit, and the ship's basic information. Perform time alignment and longitudinal projection calculations on the following ship's positioning sampling data, the leading ship's broadcast sampling data, and the following ship's motion sampling data to form a longitudinal following sample sequence of the fore-and-aft ships that has completed spatiotemporal registration.

[0023] In one embodiment of the present invention, the lead vessel is a target vessel located ahead of the navigation formation during convoy navigation in inland waterways, providing navigation status reference information to the following vessels. The lead vessel broadcasts its own positioning information (longitude, latitude) and motion status information (lead vessel heading angle data) to the following vessels via the formation data link. The following vessel is a vessel located behind the lead vessel during convoy navigation in inland waterways, performing longitudinal following control based on the lead vessel's status information and its own motion status information. The following vessel collects its own positioning, motion, and control status information (follower vessel propulsion speed and follower vessel rudder angle) through a ship-mounted RTK positioning receiver, inertial measurement unit, and propulsion controller.

[0024] Specifically, in step S11, the timestamps in the following vessel's positioning sampling data, the preceding vessel's broadcast sampling data, and the following vessel's motion sampling data are obtained, and the obtained timestamps are converted into a reference time under a unified time base. The sampling period in the ship's basic information is used as the time interval between adjacent sampling times in the unified sampling time series, wherein the reference time and the sampling time in the unified sampling time series are at a unified time base. For the unified sampling time series, the following vessel's positioning sampling data, the preceding vessel's broadcast sampling data, and the following vessel's motion sampling data under the reference time that is consistent with the sampling time are extracted as the sampling data under that sampling time. The sampling data are sorted according to the order of occurrence of the sampling time to obtain a synchronous sampling data sequence. Specifically, the planar position coordinates of the leading ship and the following ship are both 2D planar coordinates. The planar position coordinates of the leading ship include the lateral coordinates and the longitudinal coordinates of the leading ship, and the planar position coordinates of the following ship include the lateral coordinates and the longitudinal coordinates of the following ship. Specifically, the calculation formulas for the longitudinal distance between the fore and aft ships and the longitudinal relative velocity at different sampling times in step S13 are as follows: ; ; in, This represents the longitudinal distance between the preceding and following vessels at the nth sampling time. This represents the longitudinal relative velocity between the preceding and following vessels at the nth sampling time. Indicates the sampling period. This represents the lateral coordinate of the ship's front plane at the nth sampling time. This represents the longitudinal coordinate of the forward ship at the nth sampling time. This represents the lateral coordinate of the following ship in the plane at the nth sampling time. This represents the longitudinal coordinate of the following ship in the plane at the nth sampling time. This represents the heading angle data of the following ship at the nth sampling time, where N represents the total number of sampling times; The longitudinal distance between the fore and aft vessels refers to the distance obtained by projecting the position difference vector between the fore and aft vessels along the current course of the trailing vessel, which is used to represent the actual distance between the fore vessel and the trailing vessel in the course of travel; the longitudinal relative speed represents how fast the longitudinal distance between the fore and aft vessels changes.

[0025] S2: Based on the longitudinal following sample sequences of the fore and aft ships that have completed spatiotemporal registration, generate a single ship longitudinal acceleration sequence and a reference longitudinal hydrodynamic action term sequence. Calculate the residual sequence between the longitudinal acceleration sequence of the following ship and the longitudinal acceleration sequence of the single ship and the reference longitudinal hydrodynamic action term sequence in the longitudinal following sample sequences of the fore and aft ships. Use this as a longitudinal interactive residual sequence. Concatenate the longitudinal interactive residual sequence to the longitudinal following sample sequences of the fore and aft ships to obtain the longitudinal following sample sequences of the fore and aft ships that have completed the longitudinal interactive residual calibration.

[0026] Specifically, in step S21, the arithmetic mean of the ship lengths is obtained by adding the lengths of the foreship and the following ship and then dividing by two. The normalized longitudinal distance between the foreship and the following ship is obtained by dividing the longitudinal distance between the foreship and the following ship by the arithmetic mean of the ship lengths. Furthermore, in step S23, by acquiring the following ship's propulsion speed, rudder angle, and draft at different sampling times, multidimensional matching and interpolation calculations are performed on the single ship's longitudinal acceleration calibration curve to obtain the single ship's longitudinal acceleration. The longitudinal interaction residual value at the same sampling time is obtained by calculating the longitudinal acceleration of the following ship minus the longitudinal acceleration of the single ship minus the reference longitudinal hydrodynamic action term. The longitudinal interaction residual value is then sorted according to the order of occurrence of the sampling time to obtain the longitudinal interaction residual sequence.

[0027] The reference longitudinal hydrodynamic coefficient table consists of several fixed longitudinal spacing nodes with equal spacing and the reference longitudinal hydrodynamic coefficients corresponding to the fixed longitudinal spacing nodes. Two adjacent fixed longitudinal spacing nodes are used as a set of longitudinal spacing grids, and two adjacent fixed longitudinal spacing nodes are used as the left grid spacing point and the right grid spacing point of the longitudinal spacing grid, respectively, where the right grid spacing point is greater than the right grid spacing point. If the normalized longitudinal spacing lies between the left and right grid spacing points in any longitudinal spacing grid, then based on the reference longitudinal hydrodynamic coefficients corresponding to the left and right grid spacing points of that longitudinal spacing grid, a reference longitudinal hydrodynamic term corresponding to the normalized longitudinal spacing is generated, wherein the formula for generating the reference longitudinal hydrodynamic term is: ; in, Indicates normalized vertical spacing The corresponding baseline longitudinal hydrodynamic term, Indicates normalized vertical spacing The left grid spacing point of the vertical grid in which it is located. Indicates normalized vertical spacing The right grid spacing point of the vertical grid in which it is located. Indicates the left grid spacing point The corresponding reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficients table. Indicates the right grid spacing point The corresponding reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficients table. Indicates normalized vertical spacing Corresponding left grid spacing point Spacing coefficient, Indicates normalized vertical spacing Corresponding right grid spacing points Spacing coefficient; Specifically, in step S22, the normalized longitudinal spacing .

[0028] Specifically, the reference longitudinal hydrodynamic action term represents the theoretical longitudinal hydrodynamic influence generated by the coupling effect of wake, wake and pressure field between ships under the normalized longitudinal spacing between the preceding and following ships. It is used to characterize the basic longitudinal hydrodynamic action experienced by the following ship due to the presence of the preceding ship. The longitudinal interaction residual value represents the longitudinal dynamic influence remaining in the actual longitudinal motion response of the following ship after deducting the dynamic response of the individual ship and the reference longitudinal hydrodynamic action between ships. It is used to characterize the nonlinear interaction between ships and the combined effect of external disturbances that are not described by the basic dynamic model. It can be used as a compensation quantity to describe the dynamic interaction characteristics between ships.

[0029] Furthermore, the longitudinal interactive residual calibration of the preceding and following ship tracking sample sequences is uploaded to the cloud, and the cloud performs the online correction coefficient table generation operation for the longitudinal hydrodynamic action term as described in step S3.

[0030] S3: Based on the longitudinal interactive residual calibration of the fore-and-aft ship following sample sequences, the longitudinal hydrodynamic action term correction coefficients are generated in the cloud, and an online correction coefficient table for the longitudinal hydrodynamic action term is constructed.

[0031] Specifically, the formula for constructing the difference constraint matrix of the longitudinal hydrodynamic action term attenuating with the spacing in step S31 is as follows: ; ; in, Let represent the difference constraint matrix, where It is in the form of an N x N matrix, where N represents the total number of sampling times and the total number of normalized vertical spacings. This represents the nth row and the first row of the difference constraint matrix. Column matrix elements, This represents the nth difference in the sorted difference sequence; Specifically, the observations mentioned in step S32 are column vectors of length N, and the regressor is a matrix of N rows and M columns, where M represents the number of fixed longitudinal spacing nodes in the baseline longitudinal hydrodynamic coefficient table. The nth row and mth column of the regressor represents the spacing coefficient of the nth normalized longitudinal spacing at the mth fixed longitudinal spacing node after sorting. The zero-increment anchoring term includes the zero-increment anchoring weight matrix, the zero-increment anchoring matrix, and the residual observation weight matrix. The zero-increment anchoring weight matrix and the zero-increment anchoring matrix are set to identity matrices by default. The residual observation weight matrix is ​​used to describe the calculation confidence of the N normalized longitudinal spacings, and the residual observation weight matrix is ​​set to identity matrix by default. Furthermore, the solution to the weighted least squares equation is as follows: ; in, This represents the vector of correction coefficients for the longitudinal hydrodynamic action term. Indicates the regression quantity. Represents the residual observation weight matrix. Representing the observed quantity, Let A denote the zero-increment anchor weight matrix. Represents the identity matrix. Describing the L2 norm, Indicates the selection that makes The vector c that reaches its minimum value is used as the correction coefficient vector for the longitudinal hydrodynamic action term, where... It is an M-row, 1-column column vector, where M represents the total number of fixed longitudinal spacing nodes in the reference longitudinal hydrodynamic coefficient table. The fixed longitudinal spacing nodes in the reference longitudinal hydrodynamic coefficient table are sorted in ascending order by default.

[0032] It should be noted that this invention identifies the correction coefficients for the longitudinal hydrodynamic action term based on the preceding and following ship following sample sequences with completed longitudinal interaction residual calibration. This incorporates the unmodeled interactive effects caused by inter-ship flow field interference, wake coupling, and complex hydrodynamic environments during actual navigation into the baseline longitudinal hydrodynamic coefficient table, achieving adaptive correction of the baseline longitudinal hydrodynamic coefficients. Specifically, this invention sorts the sample data according to the normalized longitudinal spacing and constructs a difference constraint matrix that decays with the spacing, ensuring that the correction coefficients corresponding to adjacent spacing positions satisfy continuous variation constraints. This avoids abrupt changes in correction coefficients due to discrete sample distribution or local anomalies, improving the smoothness and physical rationality of the longitudinal hydrodynamic correction model. Furthermore, by using the longitudinal interaction residual as the observation and the spacing coefficient of the fixed longitudinal spacing node corresponding to the normalized longitudinal spacing as the regressor, a weighted least squares equation is established. This allows following state information from multiple sampling times to participate in the solution of correction coefficients, improving the stability and noise resistance of parameter identification. Furthermore, by introducing zero-increment anchoring constraints, the correction coefficients remain close to zero when effective interactive residuals are lacking, avoiding over-correction of the baseline longitudinal hydrodynamic coefficient table. The residual observation weight matrix is ​​used to adjust for differences in the reliability of different samples, improving the adaptability of parameter updates under complex navigation conditions. The final generated online correction coefficient table for longitudinal hydrodynamic effects can be directly associated with the fixed longitudinal spacing nodes in the baseline longitudinal hydrodynamic coefficient table, achieving a fusion update of the baseline longitudinal hydrodynamic coefficients and the online correction coefficients.

[0033] S4: Embed the online correction coefficient table of longitudinal hydrodynamic action terms generated in the cloud into the pre-set prediction model of the following ship at the end of the following ship to form a prediction model of the following ship that has completed the online correction of longitudinal hydrodynamic action terms.

[0034] S41: Add the online correction coefficient table of longitudinal hydrodynamic action items generated in the cloud to the baseline longitudinal hydrodynamic coefficient table cached at the end of the following ship item by item to calculate the corrected longitudinal hydrodynamic coefficient table; S42: Based on the modified longitudinal hydrodynamic coefficient table, regenerate the modified longitudinal hydrodynamic action terms and the modified longitudinal hydrodynamic action term sequence for different sampling times according to the linear interpolation method described in step S22, and write them into the pre-set follower prediction model at the end of the follower ship to generate the follower prediction model with completed online correction of the longitudinal hydrodynamic action terms.

[0035] It should be noted that this invention achieves virtual formation and spacing control of multiple ships through end-to-cloud collaboration. The ship-side is responsible for collecting navigation status data of following and leading ships, and completing spatiotemporal registration of multi-source data, longitudinal following feature extraction, and real-time propulsion control. The cloud-side generates longitudinal hydrodynamic action term correction coefficients based on the following sample sequences of the preceding and following ships uploaded by the ship-side and sends them to the ship-side. The ship-side integrates the longitudinal hydrodynamic action term correction coefficients into a pre-set prediction model of the following ships, updates the prediction capability of hydrodynamic interaction between ships, and solves the spacing maintenance optimization control problem based on the longitudinal hydrodynamic action term online-corrected prediction model of the following ships, generating propulsion control commands. This achieves collaboration between cloud-side model optimization and ship-side real-time control, improving the accuracy and stability of virtual formation spacing maintenance in complex inland waterway navigation environments.

[0036] S5: Based on the state recursion of the follower prediction model with online correction of longitudinal hydrodynamic terms, solve the quadratic programming problem of spacing maintenance with propulsion constraints, and output the follower propulsion control command for maintaining the longitudinal distance between ships stably.

[0037] Specifically, in step S51, the following ship prediction model uses the corrected longitudinal hydrodynamic action term, longitudinal distance between the fore and aft ships, longitudinal relative velocity, longitudinal acceleration of the following ship, and rudder angle of the following ship corresponding to the current control time as the current state input. Combined with the candidate propulsion speed sequence in the prediction time domain, a recursive prediction method based on a neural network model structure is adopted to gradually generate the predicted longitudinal distance between the fore and aft ships, the predicted longitudinal relative velocity, the predicted rudder angle of the following ship, and the predicted longitudinal acceleration of the following ship for the next G time steps. Based on the square of the deviation between the predicted longitudinal distance between the fore and aft ships and the target longitudinal distance, the square of the longitudinal relative velocity, and the square of the change in propulsion speed for the next G time steps, a spacing-preserving quadratic programming objective function is constructed, where the change in propulsion speed is the difference in propulsion speed between two adjacent time steps. The default setting for G is 10, and the time interval between adjacent time steps is the sampling period. In step S52, after predicting the longitudinal distance between the fore and aft ships and the longitudinal relative speed, and constructing the objective function of the spacing-preserving quadratic programming, the propulsion speed is limited and the rate of change of propulsion speed is limited during the recursive solution of the objective function, forming a constrained quadratic programming optimization problem. The propulsion speed sequence for the next G time steps that minimizes the objective function of the spacing-preserving quadratic programming is obtained as the candidate sequence of propulsion speed in the predicted time domain.

[0038] It should be noted that this invention achieves an integrated design for predicting the future motion state of a vessel and optimizing propulsion control by combining an online-corrected follower prediction model with a spacing-maintaining quadratic programming approach. Specifically, this invention utilizes the corrected longitudinal hydrodynamic action term to improve the follower prediction model's ability to characterize the effects of inter-ship flow field coupling, making the predicted longitudinal spacing and relative longitudinal velocity of the preceding and following vessels closer to the actual navigation state. Furthermore, it constructs an optimization objective by jointly using the longitudinal spacing deviation, relative longitudinal velocity, and propulsion speed changes within the prediction time domain, and combines propulsion speed limits and propulsion speed change rate limits for constraint solving. This ensures the smooth operation of the propulsion system while improving the accuracy of maintaining the target longitudinal spacing, avoiding frequent fluctuations in propulsion speed and sudden changes in control commands. By continuously updating propulsion control commands through rolling optimization, the follower vessel can respond promptly to changes in the motion of the preceding vessel and changes in the hydrodynamic environment, improving the stability, real-time performance, and robustness of longitudinal following control in inland waterway multi-vessel virtual formations.

[0039] In another embodiment of the present invention, for ships 1, 2 and 3 in a multi-ship virtual formation scenario, during the generation of the follower propulsion control command for ship 2, ship 1 is used as the lead ship and ship 2 is used as the follower ship, and the follower propulsion control command for ship 2 is generated using steps S1 to S5 of the present invention; during the generation of the follower propulsion control command for ship 3, ship 2 can be used as the lead ship and ship 3 can be used as the follower ship, and the follower propulsion control command for ship 3 can be generated using steps S1 to S5 of the present invention, thereby realizing the virtual formation of multiple ships in inland waterways and ensuring that the formation order remains unchanged during the ship's journey.

[0040] Example 2: Figure 2 This image presents a comparison of the technical effects of the end-to-cloud collaborative virtual formation and spacing maintenance method for inland waterways described in this application. The image uses a realistic aerial photograph of a narrow inland waterway for left-right comparison. The left side shows the formation navigation effect without the online correction of the longitudinal hydrodynamic effect term described in this application, while the right side shows the formation navigation effect after applying the method described in this application. In the left-side scenario, the lead vessel and the following vessel follow each other closely along the inland waterway. Due to the mismatch between the fixed ship-to-ship hydrodynamic interaction model and the actual vessel state, the following vessel's prediction of the lead vessel's wake and the longitudinal hydrodynamic effect between vessels is inaccurate, leading to a deviation in the longitudinal distance between the following vessel and the lead vessel. This manifests as unstable spacing between vessels, localized spacing compression, or short-term downward swerving. In the right-side scenario, the following vessel receives the longitudinal hydrodynamic effect term corrected by end-to-cloud collaboration and uses the correction result to solve the ship-end prediction model and spacing maintenance control. This allows the following vessel to continuously correct the longitudinal motion prediction result based on actual changes in the spacing between vessels, thereby stably maintaining the longitudinal distance with the lead vessel without changing the formation structure and target spacing settings.

[0041] Figure 2By comparing left and right sides under the same channel, the same fleet, and the same following direction, the technical effect of the method in this application compared with the fixed hydrodynamic interaction model is intuitively demonstrated: it can reduce the longitudinal distance deviation between ships, suppress short-term downsurge, and improve the stability and safety of spacing when heterogeneous inland waterway vessels are sailing in close-range virtual formation.

[0042] As another embodiment of the present invention, Figure 3 The figures show a comparison of the alignment between the predicted trajectory of the following vessel and the actual trajectory output by the following vessel prediction model described in this invention. The upper figure shows the following navigation state without online correction of the longitudinal hydrodynamic effect, where the predicted trajectory of the following vessel deviates significantly from the actual trajectory, indicating that the fixed ship-to-ship hydrodynamic interaction model cannot accurately reflect the changes in longitudinal hydrodynamic effect during actual close following. The lower figure shows the following navigation state after online correction using the method described in this invention, where the predicted trajectory of the following vessel is basically aligned with the actual trajectory, and the longitudinal distance between ships remains more stable, indicating that this application can improve the consistency between the following vessel prediction model and the actual operating state, and enhance the spacing maintenance effect of virtual formations of multiple vessels in inland waterways.

[0043] It should be noted that the terms "comprising," "including," or any other variations thereof used herein are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0044] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0045] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration, characterized in that, The method includes: S1: Collect the following ship's positioning sampling data output by the RTK positioning receiver at the following ship's end, the leading ship's broadcast sampling data broadcast by the leading ship through the formation data link, the following ship's motion sampling data output by the following ship's inertial measurement unit, and the ship's basic information. Perform time alignment and longitudinal projection calculations on the following ship's positioning sampling data, the leading ship's broadcast sampling data, and the following ship's motion sampling data to form a longitudinal following sample sequence of the fore-and-aft ships that has completed spatiotemporal registration. S2: Based on the longitudinal following sample sequences of the fore and aft ships that have completed spatiotemporal registration, generate a single ship longitudinal acceleration sequence and a reference longitudinal hydrodynamic action term sequence. Calculate the residual sequence between the longitudinal acceleration sequence of the following ship and the longitudinal acceleration sequence of the single ship and the reference longitudinal hydrodynamic action term sequence in the longitudinal following sample sequences of the fore and aft ships. Use this as the longitudinal interactive residual sequence. Concatenate the longitudinal interactive residual sequence to the longitudinal following sample sequences of the fore and aft ships to obtain the longitudinal following sample sequences of the fore and aft ships that have completed the longitudinal interactive residual calibration. S3: Based on the longitudinal interactive residual calibration of the fore and aft ship following sample sequences, the longitudinal hydrodynamic action term correction coefficients are generated in the cloud, and an online correction coefficient table for the longitudinal hydrodynamic action term is constructed. S4: Embed the online correction coefficient table of longitudinal hydrodynamic action terms generated in the cloud into the pre-set prediction model of the following ship at the end of the following ship to form a prediction model of the following ship that completes the online correction of longitudinal hydrodynamic action terms. S5: Based on the prediction model of the following ship with online correction of the longitudinal hydrodynamic action term, solve the quadratic programming problem of spacing maintenance with propulsion constraints, and output the propulsion control command of the following ship to stably maintain the longitudinal distance between the ships.

2. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration as described in claim 1, characterized in that, In step S1, the following ship positioning sampling data includes the timestamp, longitude, latitude, and heading angle data collected by the following ship's RTK positioning receiver; the leading ship broadcast sampling data includes the timestamp, longitude, latitude, heading angle data, and longitudinal velocity data broadcast by the leading ship through the formation data link; the following ship motion sampling data includes the following ship's longitudinal acceleration collected by the following ship's inertial measurement unit, the following ship's propulsion speed output by the propulsion controller, and the following ship's rudder angle; and the ship's basic information includes the length of the leading ship, the length of the following ship, the draft of the following ship, the longitudinal distance between the target and the sampling period.

3. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways with end-to-cloud collaboration as described in claim 2, characterized in that... Step S1 involves performing time alignment and longitudinal projection calculations on the following vessel's positioning sampling data, the preceding vessel's broadcast sampling data, and the following vessel's motion sampling data. It also includes: S11: Perform time base conversion and unified sampling period processing on the timestamps in the following ship positioning sampling data, the leading ship broadcast sampling data, and the following ship motion sampling data to establish a unified sampling time series, and perform time alignment on various types of sampling data based on the unified sampling time series to obtain a synchronous sampling data series; S12: Convert the latitude and longitude in the synchronous sampling data sequence to planar position coordinates under a unified coordinate reference to form a spatial position sequence of the fore and aft ships that includes sampling time, planar position coordinates of the preceding ship, and planar position coordinates of the following ship. S13: Based on the spatial position sequence of the fore and aft ships and the heading angle data of the following ship at different sampling times, calculate the longitudinal distance between the fore and aft ships at different sampling times, and combine the changes in the longitudinal distance between the fore and aft ships at adjacent sampling times to calculate the longitudinal relative speed between the fore and aft ships at different sampling times. S14: Read the longitudinal distance between the fore and aft ships, the longitudinal relative velocity, the longitudinal acceleration of the following ship, the propulsion speed of the following ship, and the rudder angle of the following ship at different sampling times, and sort the read data according to the order of occurrence of the sampling time to obtain a sorted sequence. The sorted sequence is then concatenated with the length of the fore ship, the length of the following ship, the draft of the following ship, and the longitudinal distance between the target ships to obtain the longitudinal following sample sequence of the fore and aft ships for completing the spatiotemporal registration.

4. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration as described in claim 1, characterized in that... In step S2, based on the longitudinal following sample sequences of the fore and aft ships that have completed spatiotemporal registration, a single-ship longitudinal acceleration sequence and a reference longitudinal hydrodynamic action term sequence are generated, including: S21: Read the longitudinal distance between the fore and aft ships at different sampling times in the longitudinal following sample sequence of the fore and aft ships, and normalize the longitudinal distance between the fore and aft ships based on the length of the fore ship and the length of the following ship to obtain the normalized longitudinal distance. S22: Receive the reference longitudinal hydrodynamic coefficient table sent from the cloud, perform linear interpolation on the reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficient table based on the normalized longitudinal spacing, obtain the reference longitudinal hydrodynamic action item corresponding to the normalized longitudinal spacing, sort the reference longitudinal hydrodynamic action item based on the sequence order of the normalized longitudinal spacing in the longitudinal following sample sequence of the fore and aft ships, and obtain the reference longitudinal hydrodynamic action item sequence. S23: Read the propulsion speed and rudder angle of the following ship at different sampling times in the longitudinal following sample sequence of the preceding and following ships. Use the read content and the draft value of the following ship as the input values ​​of the single ship longitudinal acceleration calibration curve preset at the end of the following ship. Extract the single ship longitudinal acceleration at the sampling time associated with the read content from the single ship longitudinal acceleration calibration curve. Sort the single ship longitudinal acceleration according to the order of occurrence of the sampling time to obtain the single ship longitudinal acceleration sequence.

5. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration as described in claim 4, characterized in that... The S22 step, which involves linear interpolating the reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficient table based on the normalized longitudinal spacing, further includes: The reference longitudinal hydrodynamic coefficient table consists of several fixed longitudinal spacing nodes with equal spacing and the reference longitudinal hydrodynamic coefficients corresponding to the fixed longitudinal spacing nodes. Two adjacent fixed longitudinal spacing nodes are used as a set of longitudinal spacing grids, and two adjacent fixed longitudinal spacing nodes are used as the left grid spacing point and the right grid spacing point of the longitudinal spacing grid, respectively, where the right grid spacing point is greater than the right grid spacing point. If the normalized longitudinal spacing lies between the left and right grid spacing points in any longitudinal spacing grid, then based on the reference longitudinal hydrodynamic coefficients corresponding to the left and right grid spacing points of that longitudinal spacing grid, a reference longitudinal hydrodynamic term corresponding to the normalized longitudinal spacing is generated, wherein the formula for generating the reference longitudinal hydrodynamic term is: ； in, Indicates normalized vertical spacing The corresponding baseline longitudinal hydrodynamic term, Indicates normalized vertical spacing The left grid spacing point of the vertical grid in which it is located. Indicates normalized vertical spacing The right grid spacing point of the vertical grid in which it is located. Indicates the left grid spacing point The corresponding reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficients table. Indicates the right grid spacing point The corresponding reference longitudinal hydrodynamic coefficients in the reference longitudinal hydrodynamic coefficients table. Indicates normalized vertical spacing Corresponding left grid spacing point Spacing coefficient, Indicates normalized vertical spacing Corresponding right grid spacing points Spacing coefficient.

6. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration as described in claim 1, characterized in that... In step S3, correction coefficients for the longitudinal hydrodynamic action term are generated, and an online correction coefficient table for the longitudinal hydrodynamic action term is constructed, including: S31: Calculate the difference between adjacent normalized longitudinal distances in the preceding and following ship following sample sequences after longitudinal interactive residual calibration is completed. Sort the data at different sampling time points in the preceding and following ship following sample sequences after longitudinal interactive residual calibration is completed in ascending order of the difference to obtain a rearranged result sequence. Based on the difference sequence formed by the difference between adjacent normalized longitudinal distances in ascending order, construct the difference constraint matrix of the longitudinal hydrodynamic action term decaying with distance. S32: Extract the longitudinal interactive residual sequence from the rearranged result sequence as the observation, and use the spacing coefficients of the left grid spacing point and the right grid spacing point corresponding to the normalized longitudinal spacing as the regressors. Solve the weighted least squares equation containing the zero-increment anchoring term and the difference constraint matrix to obtain the longitudinal hydrodynamic action term correction coefficient vector. The longitudinal hydrodynamic action term correction coefficient vector consists of the longitudinal hydrodynamic action term correction coefficients corresponding to the fixed longitudinal spacing nodes after ascending sorting in the benchmark longitudinal hydrodynamic coefficient table. S33: Write the longitudinal hydrodynamic action item correction coefficients in the longitudinal hydrodynamic action item correction coefficient vector into the fixed longitudinal spacing nodes of the standard longitudinal hydrodynamic coefficient table after ascending order, as the online correction coefficient table for the longitudinal hydrodynamic action item.

7. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration as described in claim 1, characterized in that... In step S4, the online correction coefficient table for the longitudinal hydrodynamic action term generated in the cloud is embedded into the pre-set follower prediction model at the bow of the follower vessel, including: S41: Add the online correction coefficient table of longitudinal hydrodynamic action items generated in the cloud to the baseline longitudinal hydrodynamic coefficient table cached at the end of the following ship item by item to calculate the corrected longitudinal hydrodynamic coefficient table; S42: Based on the modified longitudinal hydrodynamic coefficient table, regenerate the modified longitudinal hydrodynamic action terms and the modified longitudinal hydrodynamic action term sequence for different sampling times according to the linear interpolation method described in step S22, and write them into the pre-set follower prediction model at the end of the follower ship to generate the follower prediction model with completed online correction of the longitudinal hydrodynamic action terms.

8. The method for virtual formation and spacing maintenance of multiple vessels in inland waterways using end-to-cloud collaboration as described in claim 1, characterized in that... In step S5, the following ship prediction model with online correction of the longitudinal hydrodynamic term is used as the basis for state recursion to solve the spacing-preserving quadratic programming problem with propulsion constraints, including: S51: Using the prediction model of the following ship with online correction of the longitudinal hydrodynamic action term as the state recursive equation, the predicted longitudinal distance between the preceding and following ships and the predicted longitudinal relative speed of the following ship in the prediction time domain are generated. The propulsion speed in the prediction time domain is used as the parameter to be solved. According to the square of the deviation between the predicted longitudinal distance between the preceding and following ships and the target longitudinal distance, the square of the longitudinal relative speed and the square of the change in propulsion speed in the prediction time domain, a quadratic programming objective function for maintaining the distance is constructed. S52: Apply propulsion speed limit and propulsion speed change rate limit to the quadratic programming objective function that maintains spacing, and solve to obtain the candidate sequence of propulsion speed in the predicted time domain; S53: Extract the first propulsion speed from the candidate sequence of predicted time-domain propulsion speeds, and generate a follower propulsion control command based on the extracted propulsion speed to maintain the longitudinal distance between ships, so that the follower ships can maintain the target longitudinal distance under the condition that the longitudinal hydrodynamic action term has been corrected online.

Images