Unmanned ship power boat interactive replacement method and device and electronic equipment

Abstract

The invention provides an interactive replacement method and device for a power boat of an unmanned ship and electronic equipment, and relates to the field of data processing. According to the method, multi-modal observation flows of an unmanned ship, a small power boat and an interaction environment are collected, a candidate factor set is extracted, short-time prediction is carried out in combination with sea condition disturbance, a proximity control instruction is generated and issued by using a multi-modal interaction protocol, geometric guidance and locking verification is triggered after correction of a buoy node, and thus the unmanned ship and the small power boat are identified. And then power switching is completed according to a continuous power distribution curve, and interactive replacement of the unmanned ship power skiff is achieved in combination with a causal atlas. By implementing the technical scheme provided by the invention, high-robustness interactive replacement of the unmanned ship and the small power boat can be conveniently realized through a collaborative coping mechanism aiming at the transient docking window and the attitude difference.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to an unmanned ship power boat interactive replacement method and device and electronic equipment. BACKGROUND

[0002] In the process of executing long-distance navigation and tasks by an unmanned ship, in order to meet the endurance and power requirements in different task scenarios, external power supplement or replacement is usually needed through a power boat. However, in the docking and replacement process of the unmanned ship and the power boat, there is obvious dynamic complexity.

[0003] Firstly, the docking window is transient. Due to the influence of sea state disturbance, wind wave superposition and flow field fluctuation, the unmanned ship and the power boat are difficult to maintain a long-time relative stable interval in the approaching process, so that the time for completing attitude alignment and guided docking is extremely limited. Secondly, the attitude difference is large. The power boat has multi-dimensional disturbance such as roll, pitch and yaw under the action of different waves, and the attitude adjustment capability of the unmanned ship itself is also limited by the propulsion redundancy and inertia constraints, so that the space attitude of both parties is difficult to keep consistent at the docking moment. Thirdly, the existing control method generally relies on single-channel sensors and local feedback loops, and the real-time control reaction has a lag. When short-time disturbance or sudden deviation occurs, it is difficult to complete correction in time within the limited docking window, so as to easily cause docking failure or excessive mechanical impact.

[0004] In summary, the existing technology lacks a coordinated response mechanism for the transient docking window and attitude difference in the interactive replacement process of the unmanned ship and the power boat, and it is urgent to propose a solution capable of realizing high-robustness interactive replacement. SUMMARY

[0005] The present application provides an unmanned ship power boat interactive replacement method, which realizes high-robustness interactive replacement of the unmanned ship and the power boat through a coordinated response mechanism for the transient docking window and attitude difference.

[0006] In a first aspect of the present application, an unmanned ship powered boat interactive replacement method is provided, the method comprising: acquiring multi-modal observation streams of an unmanned ship, a powered boat, and an interactive environment, and extracting a candidate factor set from the multi-modal observation streams based on a disturbance field reconstruction pipeline; inputting the candidate factor set into a short-time trajectory prediction and attitude prediction engine to generate trajectory prediction and attitude prediction results in combination with sea state disturbance information; broadcasting the trajectory prediction and attitude prediction results to the unmanned ship and the powered boat under a multi-modal interaction protocol, and generating a proximity control instruction set according to an interactive intention sharing stream fed back by the unmanned ship and the powered boat after the broadcasting; issuing the proximity control instruction set to the unmanned ship and the powered boat, correcting the trajectory prediction and attitude prediction results with the support of disturbance correction services provided by a buoy node, outputting a disturbance-corrected proximity state stream, and triggering a docking geometric guidance process when a proximity error falls within a threshold range to complete a locking verification; after the docking locking state is confirmed, adjusting the power output of the powered boat according to a continuous power distribution curve in accordance with a contact load spectrum and a locking state, and simultaneously reducing the power output of the unmanned ship to obtain a smooth switching trajectory and a power scheduling trajectory; collecting trajectory deviation, attitude deviation, contact load, and execution feedback throughout the entire interactive process to construct a causal graph, and completing the unmanned ship powered boat interactive replacement based on the smooth switching trajectory, the power scheduling trajectory, and the causal graph.

[0007] In a second aspect of the present application, an unmanned ship powered boat interactive replacement device is provided, the device comprising an acquisition module and a processing module, wherein the acquisition module is configured to acquire multi-modal observation streams of an unmanned ship, a powered boat and an interactive environment, and extract a candidate factor set from the multi-modal observation streams based on a disturbance field reconstruction pipeline; the processing module is configured to input the candidate factor set into a short-time trajectory prediction and attitude prediction engine, combine sea state disturbance information to generate trajectory prediction and attitude prediction results; the processing module is further configured to broadcast the trajectory prediction and attitude prediction results to the unmanned ship and the powered boat under a multi-modal interaction protocol, and generate a proximity control instruction set according to an interactive intention sharing stream fed back by the unmanned ship and the powered boat after the broadcasting; the processing module is further configured to issue the proximity control instruction set to the unmanned ship and the powered boat, correct the trajectory prediction and attitude prediction results under the support of disturbance correction services provided by a buoy node, output a disturbance-corrected proximity state stream, and trigger a docking geometric guidance process when a proximity error falls within a threshold range to complete a locking verification; the processing module is further configured to, after the docking locking state is confirmed, adjust the power output of the powered boat according to a continuous power distribution curve based on a contact load spectrum and a locking state, and simultaneously reduce the power output of the unmanned ship, to obtain a smooth switching trajectory and a power scheduling trajectory; the processing module is further configured to collect trajectory deviation, attitude deviation, contact load and execution feedback throughout the entire interactive process, construct a causal graph, and complete the unmanned ship powered boat interactive replacement based on the smooth switching trajectory, the power scheduling trajectory and the causal graph.

[0008] In a third aspect of the present application, an electronic device is provided, comprising a processor, a memory, a user interface and a network interface, the memory is configured to store instructions, the user interface and the network interface are configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method described above.

[0009] In a fourth aspect of the present application, a computer readable storage medium is provided, which stores instructions that, when executed, perform the method described above.

[0010] In summary, one or more technical solutions provided in the present application have at least the following technical effects or advantages: Comprehensive perception of unmanned ship, power boat and environment is realized through multi-modal observation of flow and disturbance field reconstruction, so that prediction and control have higher comprehensiveness and accuracy. Through the combination of short-time trajectory prediction and attitude prediction engine with sea state disturbance information, future state results are generated in advance, solving the problem of short docking window and large attitude difference, and improving the forward-looking of the docking process. With the help of multi-modal interaction protocol for information broadcasting and intention sharing, unmanned ship and power boat can make collaborative decisions during docking, reducing the reaction lag of single control loop. The disturbance correction service provided by the buoy node is used to correct the trajectory prediction and attitude prediction results in real time, ensuring the docking stability under sudden disturbance, and triggering geometric guidance when the error meets the threshold, realizing accurate locking. After locking confirmation, the continuous power distribution curve is introduced, which can smoothly complete the power switching between power boat and unmanned ship, avoiding interruption and mechanical impact caused by power mutation. The trajectory deviation, attitude deviation, contact load and execution feedback are collected throughout the process, and a causal graph is constructed, combined with the smooth switching trajectory and power scheduling trajectory, so that the system has closed-loop monitoring and abnormal backtracking capability. In summary, through the collaborative response mechanism for short docking window and attitude difference, high-robustness interactive replacement of unmanned ship and power boat is facilitated. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 A flowchart of an unmanned ship and power boat interactive replacement method provided by an embodiment of the present application is shown. Figure 2 A module diagram of an unmanned ship and power boat interactive replacement device provided by an embodiment of the present application is shown. Figure 3 A structural diagram of an electronic device provided by an embodiment of the present application is shown.

[0012] Explanation of reference signs: 21, acquisition module; 22, processing module; 31, processor; 32, communication bus; 33, user interface; 34, network interface; 35, memory. DETAILED DESCRIPTION

[0013] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be described clearly and completely in conjunction with the drawings in the embodiments of the specification. Obviously, the described embodiments are only some of the embodiments of the present application, not all.

[0014] In the description of the embodiments of the present application, the words such as "for example" or "for instance" are used to indicate an example, an illustration or an illustration. Any embodiment or design scheme described as "for example" or "for instance" in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the use of words such as "for example" or "for instance" is intended to present the relevant concept in a specific manner.

[0015] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first", "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined with "first", "second" can be explicitly or implicitly included one or more features. The terms "include", "contain", "have" and their variants mean "include but not limited to", unless otherwise specifically emphasized.

[0016] In the embodiments of the present application, the working mode of the unmanned ship and the power boat includes the following aspects: the power boat has a modular design feature, which can be dynamically embedded into the tail of the unmanned ship during the task process and used as a power output unit. The power boat adopts a loading method similar to a rocket bundled booster, which can be flexibly configured according to the task demand of the unmanned ship, can be loaded individually or multiple, and has a series of model specifications, which can be matched for unmanned ships of different tonnages, thereby meeting the power demand of multiple scenarios. The unmanned ship has the ability of autonomous discovery and selection when performing tasks, and can search for a group of power boats located at a fixed position on the sea or on the shore of a port through autonomous navigation, so as to ensure that the appropriate power supply can be obtained in different operating environments. After the unmanned ship reaches the group of power boats, interactive cooperative operation can be performed between the two parties to realize autonomous replacement and loading of the power boats, thereby ensuring the flexibility of power scheduling and the continuity of power supply of the unmanned ship during the task process.

[0017] The present application provides an interactive replacement method of unmanned ship power boat, referring to Figure 1 , Figure 1 A flowchart of an interactive replacement method of unmanned ship power boat provided by the embodiments of the present application. The method is applied to a server and includes steps S110 to S160, and the above steps are as follows: S110, acquire multi-modal observation streams of the unmanned ship, the power boat and the interactive environment, and extract a candidate factor set from the multi-modal observation streams based on a perturbation field reconstruction pipeline.

[0018] Specifically, server refers to computing and communication resources deployed on the buoy node, shore base station or shipborne computing unit, for unified access, processing and distribution of data, supporting distributed soft bus and deterministic scheduling channel. For example, the industrial-grade server of the shore base station subscribes to the video channel and acoustic positioning channel of the buoy node through the distributed soft bus, while exchanging attitude state and propulsion state with the shipborne computing unit for online inference and instruction issuance in joint operation. Unmanned ship refers to an unmanned ship entity with autonomous navigation, perception and execution capabilities, as a controlled object for docking and replacement process, continuously outputting navigation state, attitude state and propulsion state. For example, a 100-ton unmanned ship outputs relative distance, relative orientation, yaw angular velocity and propulsion redundancy label in the approach stage, for guiding approach corridor correction and attitude target band correction. Power boat refers to a small power carrier with modular connection interface and independent propulsion output capability, which can participate in continuous power distribution as a power output unit of the unmanned ship after locking. For example, the series of power boats provide different thrust levels and health status labels, and participate in load balancing according to power scheduling trajectories when concurrent access.

[0019] Interaction environment refers to a set of dynamic external conditions around the docking area of the unmanned ship and the power boat, including sea state disturbance, wind field change, flow field shear, adjacent obstacle and communication link state. For example, there are surge superposition and transverse nearshore flow outside the port, which shortens the docking window and introduces attitude difference amplification. Multi-modal observation stream refers to a set of heterogeneous sequences continuously generated by multiple types of sensors under a unified time reference and time-aligned, covering position information, velocity information, acceleration information, attitude information, sea state disturbance information and structural response information. For example, the vision channel outputs video frames and target boxes, the acoustic positioning channel outputs time difference of arrival and positioning accuracy, the inertial navigation unit outputs three-axis angular velocity and three-axis linear acceleration, the satellite and short baseline joint positioning component outputs high-precision pose, the wave spectrum measurement unit outputs wave direction and main period label, and the propulsion component strain gauge and temperature sensor output load and temperature rise trajectory. All channels enter the distributed soft bus to form a continuous sequence with a unified timestamp.

[0020] The disturbance field refers to the spatiotemporal disturbance description formed by the combined action of surges, swell superposition, local eddies, wind gusts, and flow field shear within the docking space-time range, serving as the boundary condition setting for prediction and control. For example, the swell propagates from outside the harbor to the inner harbor and reflects at the revetment, superimposes to form an energy band in a certain direction, affecting the geometric stability zone of the approach corridor. The disturbance field reconstruction pipeline refers to the processing flow for generating disturbance element sequences and scene labels that can be used for decision-making from multi-modal observation streams, including cross-domain alignment, noise suppression, coordinate registration, event-based segmentation, short-time window fusion, and boundary condition injection. For example, the buoy node side completes cross-domain alignment and noise suppression of vision and wave spectrum, and the shore base station side injects far-field wind wave forecast and tidal boundary, and jointly outputs wave direction labels, energy density labels, and vortex occurrence probability labels in chronological order.

[0021] The candidate factor set refers to a standardized feature set highly relevant to approach planning, attitude control, and continuous power allocation and directly usable by the executable interface, with source identification, time window length, sampling frequency, phase label, and execution timeliness label, filtered from multi-modal observation streams under the constraint of the disturbance field reconstruction pipeline. For example, the attitude coupling factor includes relative heading angle deviation, yaw angular velocity, pitch change rate, and attitude target band occupancy rate, the propulsion coupling factor includes thrust amplitude label, thrust pulsation amplitude, propulsion redundancy ratio, and contact load trend, and the environmental factor includes swell direction label and wave group main period label, which together constitute the candidate factor set.

[0022] Further, around sensor access and time base unification, the data acquisition process of the distributed soft bus is started, and position information, velocity information, acceleration information, attitude information, sea state disturbance information, and structural response information are continuously collected, and each channel data is mapped to a unified time reference, and then time domain interpolation and missing data completion are performed, and a multi-modal observation stream under consistent sampling constraints is output. To achieve unified time reference and smooth sequence, clock offset correction, linear interpolation, and exponential smoothing are performed in sequence, as follows: ; ; ;

[0023] wherein, is the original timestamp; is the estimated global clock offset; is the unified timestamp; is the observation quantity of any channel at the unified time; is the unified timestamp of adjacent samples; is the smoothed observation quantity; is the smoothing factor; is the sampling period. The heterogeneous channels are aligned to a uniform time base by the first equation, continuous trajectories are generated between sampling points by the second equation, and high-frequency noise is suppressed and slight missing data is stably completed by the third equation.

[0024] After obtaining the multi-modal observation stream, cross-domain alignment and noise suppression are performed, coordinate registration and observation denoising are completed, and the purified multi-modal observation stream enhancement result is output. First, the observation of each channel is mapped to a unified spatial reference frame, and then the time series is denoised using state estimation, and the core relationship is as follows: ; ;

[0025] wherein, is the position or attitude vector in the sensor coordinate system; is the corresponding vector in the reference coordinate system; is the rotation matrix; is the translation vector; and are the predicted and corrected state estimations at time ; is the observation vector at time ; is the observation matrix; is the gain matrix; is the prediction covariance; is the observation noise covariance. Cross-vehicle coordinate alignment is achieved by the first equation, and noise is suppressed and observation consistency is improved in the time series dimension by the second equation, thereby obtaining an enhanced result that can be used for event identification.

[0026] Based on the multi-modal observation stream enhancement result, event processing is performed, the continuous sequence is cut into data segments organized in short time windows, and track change events, attitude mutation events, surge impact events, roll amplification events and propulsion load sudden increase events are labeled, and an intermediate clue set is output. The event labeling uses a joint threshold of standard deviation and change rate in a sliding window, and the detection quantity and triggering criterion are defined as follows:

[0027] wherein, is the feature value at the center of the window; are the mean and standard deviation in the window, respectively; is the standard deviation; is the first-order difference; is the deviation threshold; is the change rate threshold; is the event indicator. The above discrimination is run in parallel in each channel to ensure timely capture of short-term disturbances and sudden abnormalities, and the sequence is cut into intermediate clue segments with event boundaries.

[0028] The intermediate clue set is input into the disturbance field reconstruction pipeline, which fuses local sea state observations, adjacent power boat broadcasts, historical experience cache scene labels, and remote field wave forecast information to generate a disturbance element sequence containing surges, swell superposition, and local eddies, and attach occurrence probability labels, duration labels, and spatial impact range labels to each disturbance element. The multi-source fusion uses a posterior consistency framework to determine the most likely state of the disturbance element, and the core estimation uses the following formula:

[0029] wherein, is the discrete or continuous state set of the disturbance element (such as wave direction, energy density, eddy intensity); respectively represent local observations, adjacent broadcasts, historical scene labels, and remote forecasts; is the likelihood function; is the prior; is the posterior distribution after fusion; is the most likely disturbance element at time . Through this framework, near-field and far-field information are unified under the same spatiotemporal semantics to form a time-ordered disturbance element sequence.

[0030] Based on the disturbance element sequence, constraint aggregation is performed on the channel features in the multi-modal observation stream enhancement result to extract attitude coupling factors related to roll, pitch, and yaw, and propulsion coupling factors related to thrust amplitude, thrust direction, and thrust pulsation. Robustness testing and redundancy compression are performed on both types of factors, and finally a candidate factor set is output. To ensure the strong relevance and executability of the factors to the docking task, a mutual information screening, robust weighting, and sparse compression joint mechanism is adopted, defined as follows: ; ;

[0031] wherein, is the th channel feature; is the task-related target (attitude stability index or propulsion stability index); is the mutual information; is the screening threshold; is the factor index set selected by mutual information screening; is the factor residual; is the Huber-type influence function; is the robust turning point; is the robust weight; is the factor matrix selected by screening; is the target vector; is the compression coefficient; The sparse regular coefficients are used to reserve information-contributing factors by the first formula, suppress outlier impact when abnormal segments exist by the second formula, eliminate redundancy and improve interpretability by the third formula, and obtain a candidate factor set with source identification, time window length, sampling frequency, phase label, and execution time label, which is used as the standardized input of the subsequent short-time trajectory prediction and attitude prediction engine.

[0032] In S120, the candidate factor set is input into the short-time trajectory prediction and attitude prediction engine to generate trajectory prediction and attitude prediction results in combination with sea state disturbance information.

[0033] Specifically, the short-time trajectory prediction and attitude prediction engine refers to a prediction component set that outputs the trajectory segment sequence and attitude segment sequence of the unmanned ship and the powerboat for the near-future time window, which usually includes a state management unit, a channel alignment and weight constraint unit, an external disturbance interface, a multi-scenario unfolding unit, a constraint screening unit, and an uncertainty evaluation unit, and continuously outputs results in a rolling manner. For example, after the engine receives the candidate factor sequence package and the sea state disturbance information in the current rolling prediction period, it generates a relative displacement trajectory and a relative attitude sequence on the scale of several seconds, and simultaneously gives a confidence label. Short-time refers to a limited time window in the near future, emphasizing the prediction coverage of the control interval that can be directly executed within the docking window, taking into account real-time performance and operability. For example, a time window of several seconds can capture rapid attitude changes caused by swells, and also facilitate the landing execution of a close control instruction set within the same effective period. Sea state disturbance information refers to a disturbance factor set obtained by fusing local sea state observations, adjacent powerboat reports, historical experience cache scene labels, and far-field wave forecast, which usually includes the occurrence probability label, duration label, and spatial influence range label of swells, swell superposition, and local vortex, and is used to modify the engine's expectations for short-time dynamics. For example, when the far-field forecast shows that the energy of a specific wave direction is enhanced and the local observation confirms the short-time arrival, the engine will place this disturbance scenario in a high-weight branch when unfolding multiple scenarios.

[0034] Trajectory prediction refers to the forward-looking description of the spatial position sequence and relative displacement sequence of the unmanned ship and the powered boat within a short time window, which is usually organized as a sequence of trajectory segments with discrete time indices, and each segment is attached with a time-to-completion label and a confidence label, which are used to generate proximity corridor correction instructions and speed constraint instructions. For example, if the trajectory prediction shows that the relative lateral deviation will converge quickly in the future several sampling periods, then the subsequent proximity control instruction set can reduce the safety buffer and increase the upper limit of the fitting speed. Attitude prediction refers to the forward-looking description of the attitude angle sequence and angular velocity sequence of the unmanned ship and the powered boat within a short time window, covering attitude dimensions such as yaw, pitch and roll, and also attached with a time-to-completion label and a confidence label, which are used to generate attitude target band correction instructions and phase alignment instructions. For example, if the attitude prediction shows that the pitch will have a short-term negative pulse, then the instruction set allocates attitude execution margin in advance to suppress the pulse and avoid geometric mismatch at the moment of docking.

[0035] The trajectory prediction and attitude prediction result set consists of the main trajectory and the main attitude produced by the current rolling prediction period, as well as a set of backup trajectories and backup attitudes, and contains confidence labels, disturbance scenario labels and phase labels, which are used as the payload of the multi-modal interaction protocol broadcast and directly drive the integrated generation of proximity control instruction set. For example, the main trajectory and the main attitude in the result set are used for the execution path of the next control period, and the backup trajectory and the backup attitude are quickly switched when abnormal signs are found in the causal graph monitoring to ensure continuous controllability within the docking window.

[0036] Further, after performing missing data completion, anomaly mutation smoothing and long-term drift correction on the candidate factor set, a feature sequence package is formed and written into the short-time trajectory prediction and attitude prediction engine to establish the context of the current rolling prediction period. Missing data completion is linearly interpolated between adjacent valid samples, anomaly mutation smoothing limits the mutation amplitude with a saturation increment, and long-term drift correction removes slow drift with a sliding mean, and then a context vector is obtained by weighted aggregation for engine state initialization, which is specifically as follows: ; ; ;

[0037] wherein, is the original observation of any channel; is the adjacent valid timestamp; is the interpolation result, which obtains continuous estimation by linear proportion of the adjacent two points; is the sequence of the first frame observation; is the smoothing result; is the smoothing step; is an amplitude clipping function, the output of which is limited to to suppress mutations; is the de-drifted sample; is the sliding window length; is the multi-channel de-drift vector; is the channel-level feature map; is the time decay weight satisfying ; is the context of the current rolling prediction horizon. The first equation generates continuous values within the missing gap, the second equation limits the spike impact by saturating the increment, and the third equation eliminates slow drift by local mean and forms the context. The output feature sequence is jointly fed into the engine.

[0038] Based on the disturbance factor sequence and the above context, a multi-scenario candidate trajectory cluster and a multi-scenario candidate pose cluster are generated, and reachability filtering is performed under the constraint set consisting of geometric shape constraints, propulsion member constraints, docking guide constraints, and buffer member constraints. For each disturbance scenario the state is recursively propagated within the prediction horizon length to form the candidate, and the constraint inequalities are used to filter out unqualified trajectories and poses. The process is represented as: ; ;

[0039] where, is the joint state at time under scenario , which includes the relative position trajectory and the pose sequence ; is the control candidate; is the rolling evolution map under scenario , which is adjusted by to keep consistent with the context; is the geometric alignment error metric; is the velocity vector; is the angular velocity vector; is the contact load estimation function; is the corresponding threshold. The first equation unfolds the candidate trajectory and pose under a given scenario, and the second equation uses four types of hard constraints to limit geometry, velocity, angular velocity, and load, eliminating unexecutable candidates and retaining a set that satisfies the execution reachability.

[0040] The uncertainty of the candidate set satisfying the execution reachability is evaluated and prioritized, and then the main trajectory and the main attitude are selected as the trajectory prediction and attitude prediction results of this period, and several backups are registered for rollback. Uncertainty is jointly evaluated by integrated variance and residual consistency, stage priority is weighted according to different fault tolerance strategies of approaching stage, fitting stage and switching stage, and finally the optimal is selected by comprehensive index, and the calculation relationship is: ; ;

[0041] Wherein, the index traverses the candidate; is the residual vector of the candidate on the historical similar segment; is the integrated uncertainty covariance of the candidate ; is the confidence label, the larger the value, the lower the uncertainty and the higher the consistency with history; is the average geometric alignment error of the candidate in the predicted horizon; is the smoothness index, which is used to punish the excessive fluctuation of speed and angular velocity; is the risk index, which is used to measure the critical occupation of contact load threshold and safety buffer occupation; is the stage-related weight; is the comprehensive cost; is the main candidate index. The first formula obtains the confidence of data consistency based on residual and covariance, the second formula jointly considers stage priority, smoothness, risk and confidence in optimization, and the third formula selects the main trajectory and the main attitude and naturally connects to the subsequent multi-modal interaction protocol broadcast and proximity control instruction set generation link.

[0042] S130, broadcast the trajectory prediction and attitude prediction results to the unmanned ship and the power boat under the multi-modal interaction protocol, and generate the proximity control instruction set according to the interactive intention sharing flow fed back by the unmanned ship and the power boat after broadcasting.

[0043] Specifically, the multi-modal interaction protocol refers to the data representation and communication protocol agreed upon on the distributed soft bus, supporting signed authentication, deterministic and ordered delivery, and retransmission in case of failure, capable of simultaneously carrying heterogeneous loads such as trajectory segment sequences, attitude segment sequences, disturbance scenario labels, and execution time labels, while maintaining semantic consistency and time consistency. For example, a single data packet contains a rolling prediction period identifier, a source identifier, a main trajectory and a main attitude, and a backup trajectory and a backup attitude, and is synchronously distributed to multiple end nodes in a deterministic and ordered manner. Broadcasting refers to sending the same data packet to multiple receiving ends at the same effective time window according to the multi-modal interaction protocol, and ensuring that the receiving ends are reorganized and checked consistently according to the unified timestamp, thereby avoiding control divergence due to differences in receiving order. For example, the server issues the same main trajectory and main attitude of a rolling prediction period to the unmanned ship and the power boat, and after local authentication, they enter the same pre-execution state.

[0044] The interactive intent sharing stream refers to a standardized sequence of intent information generated and continuously updated by the unmanned ship and the power boat, used to express their respective preferences and constraints for trajectory adoption, attitude correction, and propulsion allocation, and is checked for semantic consistency and conflict alignment by the intent aggregator to form a complete and usable collaborative constraint view. For example, if the unmanned ship prefers to reduce lateral deviation and the power boat prefers to reduce transient load, the aligned intent sharing stream gives a safe buffer recommendation compatible with both. Generation refers to the intent aggregator synthesizing hard constraints and soft constraints according to priority after fusing trajectory prediction and attitude prediction results and the interactive intent sharing stream, and synthesizing an executable control scheme within the boundaries of reachability, continuity, and timeliness, forming a proximity control instruction set for both parties. For example, geometric alignment, speed range, angular velocity range, and load threshold are met first, and then smoothness and energy consumption are optimized within the propulsion redundancy and attitude adjustment margin.

[0045] The proximity control instruction set refers to the executable instruction set obtained in the above generation process, which is issued to the unmanned ship and the power boat respectively and matches the execution interface of the instruction, usually containing speed constraint instructions, yaw constraint instructions, pitch constraint instructions, proximity corridor correction instructions, attitude target band correction instructions, and safety buffer setting instructions, and is accompanied by execution time labels and rollback priority labels to ensure synchronous effectiveness and abnormal rollback within a short time window. For example, the instruction set requires reducing the upper limit of lateral speed, tightening the proximity corridor width, and increasing the attitude maintenance weight within the next effective period, while preloading backup instructions for rapid switching in the presence of sudden swells.

[0046] Further, the trajectory prediction and the posture prediction result are encapsulated into a data packet according to a multi-modal interaction protocol, the data packet payload includes a uniform timestamp, a rolling prediction period identifier, a trajectory segment sequence, a posture segment sequence, a disturbance scenario label, a confidence label, and an execution time limit label, and is attached with an authentication fingerprint and a global ordered key, and is sent to the unmanned ship and the power boat in a deterministic and ordered broadcast manner through a distributed soft bus. In order to ensure authentication and timing consistency, fingerprint and ordered key calculation is adopted, as follows: ; ; ;

[0047] wherein, represents the data packet payload; represents the symmetric key; represents the authentication fingerprint; represents the uniform timestamp (milliseconds); represents the rolling prediction period identifier; represents the broadcast ordered key; represents the number of bits used for splicing and sorting; represents the local sequence number of the receiving end; represents the time validity deviation; represents the current time of the receiving end; represents the upper limit of the allowed time deviation. The fingerprint check prevents content tampering, the ordered key ensures that multiple ends are reorganized in the same order, and the time deviation constraint ensures the execution time limit.

[0048] After the unmanned ship and the power boat receive the data packet, an interactive intention sharing stream preview frame is generated according to the local security policy, the preview frame includes trajectory adoption weight, posture adoption weight, propulsion redundancy, posture adjustment margin, energy margin, and health status label. In order to make the adoption weight interpretable and consistent with the local boundary, a normalization mapping based on confidence and boundary margin is adopted as follows:

[0049] wherein, and represent the trajectory adoption weight and the posture adoption weight, respectively; represents the Sigmoid mapping; represents the weight shaping coefficient; represent the confidence labels of the trajectory and the posture, respectively; represents the occupation margin of the upper bound of the speed; Occupancy margin representing the upper bound of angular velocity. Propulsion redundancy represents the percentage of thrust margin available for the approach phase; attitude adjustment margin represents the remaining control amount available for roll, pitch, yaw compensation; energy margin represents the percentage of available energy; health status label represents the health level of propulsion components and attitude execution units.

[0050] The interactive intent sharing stream frames from the unmanned ship and the powered boat are converged at the intent aggregator, and semantic consistency checks and conflict alignment are performed to generate a conflict summary; when there are directional conflicts, timing conflicts or resource conflicts, supplementary attitude target band priorities and propulsion component temperature rise limit constraints are requested to form a complete interactive intent sharing stream. The conflict metric adopts a weighted combination of direction, timing and resource occupation, wherein the direction deviation, timing deviation, resource gap ratio and comprehensive conflict score are calculated by the following formulas: ; ;

[0051] Wherein, and represent the expected approach direction vector of the unmanned ship and the powered boat respectively; represents the direction deviation; represents the uniform timestamp of the opponent data; represents the timing deviation; represents the required propulsion capacity; represents the current available propulsion capacity; represents the resource gap ratio; represents the weighting coefficient; represents the comprehensive conflict score. The attitude target band priority is used to determine the execution order of the attitude dimension; the propulsion component temperature rise limit constraint is used to limit the short-term power change rate to avoid overheating.

[0052] The complete interactive intent sharing stream is superimposed with the trajectory prediction and attitude prediction results, and the sea state disturbance scenario label, unmanned ship constraints and powered boat constraints to form a priority-ordered constraint sequence, and according to which a set of approach control instructions is generated, including speed constraint instructions, yaw constraint instructions, pitch constraint instructions, approach corridor correction instructions, attitude target band correction instructions and safety buffer setting instructions. The constraint sorting is performed according to the rule that hard constraints are given priority and soft constraints are given priority, while considering the time limit and safety margin, and using priority score:

[0053] Wherein, represents the priority score of the th constraint; Indicates hard constraint trigger flag (geometric alignment, velocity range, angular velocity range, and anti-indication of load threshold satisfaction); Indicates time urgency (normalized countdown to the end of the validity window); Indicates safety relevance (sensitivity to safety buffer and temperature rise limit); Indicates the weighting coefficient. When generating various types of instructions, projection and clipping operations are used for speed upper limit, angular velocity upper limit, and corridor boundary to make the target value fall within the current reachable domain; for attitude target band correction and safety buffer setting, the minimum adjustment amount strategy is adopted according to priority to ensure that the gradual correction can be verified within the execution time limit, and each instruction is attached with an execution time label and a rollback priority label to support abnormal fast switching.

[0054] S140, the approaching control instruction set is issued to the unmanned ship and the power boat, and the trajectory prediction and attitude prediction results are corrected under the disturbance correction service provided by the buoy node, the approaching state flow corrected by disturbance is output, and the docking geometric guidance flow is triggered when the approaching error falls within the threshold range to complete the locking verification.

[0055] Specifically, the buoy node is an edge computing and communication unit deployed at a fixed position in the sea area, continuously collects sea state disturbance information, and provides low-latency data services to surrounding devices, and can carry disturbance correction services and temporary cache. For example, the buoy node publishes near-field wave direction and main cycle label for quick correction of prediction in the approaching stage. The disturbance correction service refers to an online service based on the near-field sea state observation of the buoy node and the historical disturbance cache, which performs short-time correction on the trajectory prediction and attitude prediction results, and outputs the corrected segments to improve the consistency of prediction and actual sea state. For example, after detecting a sudden transverse swell, the service immediately pushes down the correction segment for re-converging the approaching corridor.

[0056] The disturbance-corrected approaching state flow refers to a time-series state set obtained by fusing the latest correction segment with the local navigation state, attitude state, and propulsion state under the action of the disturbance correction service, including geometric alignment state, velocity and angular velocity boundary occupancy rate, contact expected time, and execution confidence label. For example, the state flow indicates that "it is expected to enter the contactable window in 0.8 seconds, and the boundary occupancy rate is below the threshold value". The approaching error refers to the difference between the current approaching state and the approaching target, which is usually composed of geometric alignment error, velocity deviation, and angular velocity deviation, and is used to determine whether the conditions for executing docking guidance are met. For example, when the lateral deviation, relative heading angle deviation, and relative velocity deviation are reduced to within the safety band at the same time, it is determined that the approaching error meets the conditions.

[0057] The threshold range refers to a set of upper and lower limits for quick determination and triggering, covering boundary conditions such as geometric alignment, speed range, angular velocity range, and contact load expectation, ensuring that the docking action is within the safe and reachable domain. For example, when the lateral deviation is less than the preset value and the angular velocity is less than the upper limit, enter the guidance process. The docking geometric guidance process refers to the execution sequence of low-speed fitting, lateral fine-tuning, and longitudinal fine-tuning for achieving controlled contact between the docking guide and the buffer component after meeting the threshold range, outputting docking guidance instructions and buffer configuration plans. For example, gradually reduce the lateral deviation and lock the relative orientation, so that the guide cone enters the controlled contact interval. Lock verification refers to the step of confirming the completion and reliability of the lock actuator after controlled contact occurs, determining the effectiveness of the lock through a combination of stroke confirmation, reverse pull-off test, and health status label, generating lock status and effectiveness level. For example, if the lock stroke reaches the nominal value and the reverse pull-off force is within the allowed range, the lock is confirmed to be effective and the start condition for subsequent power switching is issued.

[0058] Further, the proximity control instruction set is issued to the respective execution interfaces of the unmanned ship and the power boat, and the execution interfaces complete local atomic switching according to the unified effective time and execution time label, and generate execution receipts and safety check results as subsequent correction inputs. To ensure atomic switching and synchronous effectiveness of instructions, time window enablement and safety check determination are used, as follows:

[0059] wherein, is the actual execution instruction stream; is the instruction vector of the issued proximity control instruction set at time ; is the unified effective time; is the time length corresponding to the execution time label; is the indication function; is the local safety check result, indicates that it is passed through the check, indicates that it is rejected for execution. The execution receipt contains state marks such as "loaded", "effective", "rejected", and local timestamps, which are used to drive the time alignment of subsequent disturbance correction services.

[0060] Based on the execution receipt and safety check result, the disturbance correction service of the buoy node is called to perform error alignment and short-time correction on the trajectory segment sequence and the attitude segment sequence, obtaining a corrected segment set, and outputting the corrected trajectory segment and the corrected attitude segment as direct inputs for state fusion. Time alignment uses time shift estimation with minimum residual, followed by amplitude correction, as follows: ; ; ;

[0061] wherein, and are relative position observation and attitude observation obtained under the assistance of near-field sea state observation channel and historical disturbance cache, respectively; and are trajectory segment sequence and attitude segment sequence, respectively; is optimal time alignment offset; and are revised trajectory segment and revised attitude segment, respectively; is revision gain. Error alignment ensures time consistency, and amplitude revision suppresses short-time deviation caused by swell and eddy.

[0062] The revised trajectory segment and the revised attitude segment are fused with the navigation state, the attitude state and the propulsion state of the unmanned ship and the power boat to output the disturbance-corrected approach state flow; the geometric alignment state, the speed boundary and the angular velocity boundary are detected according to the approach state flow as a criterion, and the threshold condition is met to trigger the docking geometric guidance process. The fusion adopts a weighted consistency strategy, and the threshold determination adopts a norm constraint, which is as follows: ; ; ;

[0063] wherein, is the disturbance-corrected approach state flow; is the relative position and attitude component of the navigation state; is the estimated position and attitude compensation component associated with the propulsion state; is the fusion weight; is the relative position vector; is the reference position of the docking target pose; and are relative speed and relative angular velocity, respectively; is the threshold value; indicates that the trigger condition is met. The revised segment output in the previous paragraph is directly used as the fusion input in this paragraph, and the input-output chain is continuous.

[0064] In the docking geometric guidance process, the disturbance-corrected approach state flow is taken as the input to generate docking guidance instructions and buffer configuration plans, which are sent to the guidance execution unit of the unmanned ship and the power boat, and low-speed fitting, lateral fine adjustment and longitudinal fine adjustment are performed in turn, so that the docking guide and the buffer component enter the controlled contact interval. The guidance instruction adopts projection and gain limiting to ensure reachability and stability:

[0065] wherein, are the velocity and attitude domain guidance commands respectively; are the velocity and attitude domain guidance commands respectively; is the guidance gain; are the velocity and attitude domain guidance commands respectively; are the docking corridor velocity and attitude target bands projection operators, ensuring the commands fall into the allowed domain; are the velocity and attitude domain guidance commands respectively; are the desired relative velocity and attitude respectively; is the relative attitude vector. The trigger flag of the previous segment output is the precondition of this segment execution, satisfying the chain dependency.

[0066] During the execution of low-speed fitting, lateral fine-tuning and longitudinal fine-tuning, the contact detection sub-process is called to determine the contact distribution and contact consistency. If the preset conditions are met, the locking verification is performed and the contact load spectrum and locking state are output as the starting conditions for subsequent power switching. Contact consistency and locking verification use three types of indicators: contact coverage, load spectrum and locking stroke, as follows: ; ;

[0067] wherein, is the contact distribution consistency indicator; is the number of contacts; is the contact load time series of the th contact; is the allowable load range of the contact; is the contact load spectrum ( is the Fourier transform); is the locking effectiveness determination; is the stroke of the locking actuator; is the minimum effective stroke; is the reverse pull-out test load; is the minimum qualified pull-out force. When is not less than the threshold value and the locking determination is satisfied, the contact load spectrum and locking state are output as the input for subsequent continuous power distribution, realizing the closed-loop connection from guidance to locking.

[0068] S150, after confirming the docking locking state, according to the contact load spectrum and locking state, the power output of the power dinghy is adjusted according to the continuous power distribution curve, and the power output of the unmanned ship is simultaneously reduced, to obtain a smooth switching trajectory and a power scheduling trajectory.

[0069] Specifically, the docking lock status confirmation refers to determining the effectiveness of the mechanical lock after the controlled contact between the docking guide and the buffer member, according to the stroke confirmation signal of the lock actuator, the reverse pull-off test result, and the health status label, and giving the effectiveness level and timestamp for the starting condition of the subsequent power switching. For example, when the stroke reaches the nominal value and the reverse pull-off force is within the allowed range, it is marked as effective and enters the power switching preparation. The lock status refers to the effectiveness of the mechanical lock at the current time, including effective, partially effective, and ineffective, etc. levels, and is accompanied by constraints such as residual gap, rebound displacement, and allowable load range, which are used to limit the slope and amplitude of the subsequent continuous power distribution curve. For example, the partially effective state only allows small amplitude thrust changes and requires higher safety buffer to be maintained.

[0070] The contact load spectrum refers to the energy distribution and dominant frequency band structure of the contact load measured by the contact array over time during the contact process, which is used to identify the buffer attenuation zone, the lock stable zone, and the load fluctuation zone, and is used as the boundary input for power scheduling and attitude compensation. For example, when the load spectrum shows that the high-frequency component attenuation is complete and the low-frequency component is stable, it is determined that the upper limit of the power change rate can be increased. The continuous power distribution curve refers to a continuous, derivable, and slope-limited time sequence planning for the power set value of each propulsion channel within the switching time window, so that the total thrust and total torque remain within the target constraints during the transition period, and mechanical impact and attitude disturbance caused by instantaneous transition are avoided. For example, a slow start segment, a linear segment, and a slow stop segment are planned for the power boat, and are paired with the reverse curve of the unmanned ship propulsion channel to maintain the balance of the resultant torque.

[0071] The power output of the power boat refers to increasing the power set value of the power boat propulsion channel according to the continuous power distribution curve under the constraints of the lock state and the contact load spectrum, and checking the temperature rise, vibration, and health status label in each rolling sampling period to ensure that the power-up process is achievable and safe. For example, the thrust is increased from 30% to 70% while keeping the side force moment within limits. The power output of the unmanned ship is simultaneously reduced, which means that the power set value of the original propulsion channel of the unmanned ship is reduced according to the mirror or matching continuous power distribution curve, and the longitudinal and lateral torque is maintained to avoid sudden drop of total thrust or secondary attitude overshoot. For example, the propulsion power is smoothly reduced from 60% to 20%, and the resultant thrust direction is kept stable with phase consistency.

[0072] The smooth switching trajectory refers to a target trajectory set reflecting the evolution of the system's expected state over time during the transition from "unmanned ship main propulsion" to "power boat main propulsion", covering the size of the combined thrust, the direction of the combined thrust, the attitude compensation amount, and the safety buffer occupancy rate, and meeting the continuity and smoothness requirements. For example, the size of the combined thrust remains constant, the direction deviation gradually converges to the target bearing, and the attitude compensation amount decreases according to the plan. The power scheduling trajectory refers to the power setting value schedule for each propulsion channel within the switching window, including the target power, allowable slope, phase reference, and torque balance reference for each channel, used to drive the execution layer to update the setting value periodically and real-time correction. For example, in the multi-boat concurrent scenario, different power ramp-up curves and phase offsets are issued to each power boat to suppress thrust pulsation.

[0073] Further, the contact load spectrum is analyzed into a lock-in stable zone, a buffer attenuation zone, and a load fluctuation zone, and a switching preparation order carrying the upper limit of the thrust change rate and the upper limit of the torque change rate is generated: the contact load time series Time-frequency analysis is performed to obtain the power spectral density The energy proportion is calculated on the high-frequency band and the low-frequency band, and the upper limit of the rate is calculated according to the partition state and the lock-in state, the formula is: ; ; ;

[0074] Wherein, and are the high-frequency energy and the low-frequency energy, respectively; is the high-low frequency demarcation; is the high-frequency proportion, used to distinguish the load fluctuation zone ( high), the buffer attenuation zone ( medium and decreasing), and the lock-in stable zone ( low and low); is the load variance, is the statistical window; is the upper limit of the thrust change rate, is the upper limit of the torque change rate; is the shaping coefficient; is the lock-in effectiveness coefficient; is the allowable contact load and the allowable contact torque, is the nominal calibration value. According to the joint threshold of and , the lock-in stable zone, the buffer attenuation zone, and the load fluctuation zone label are generated, and and Write a switch preparation order as the slope and amplitude boundary of the continuous power allocation curve.

[0075] Under the constraint of the switch preparation order, the rising curve of the power output of the powerboat and the falling curve of the power output of the unmanned ship are instantiated, and the phase reference trajectory and the torque balance reference trajectory are generated to form the curve prototype set: the power setting is continuously planned with a three-segment smooth S-curve, and the paired phase reference and torque balance reference are generated with the goal of zero synthetic torque, formula: ; ; ; ;

[0076] Where, is a smooth normalized progress function; is a time normalized variable, is the switch start time, is the planned power rising and falling time; and are the power setting curves of the powerboat and the unmanned ship respectively, is the starting power, is the switch power amplitude; is the thrust vector, is a unit direction vector generated according to the reference phase; is the phase reference trajectory; is the position vector of the thrust point relative to the center of mass; is the torque balance reference trajectory, the goal is to maintain the synthetic torque balance. The time derivatives of and are subjected to the slope constraints of and to obtain the curve prototype set that satisfies the switch preparation order.

[0077] The curve prototype set is semantically aligned with the powerboat propulsion component parameters, the unmanned ship propulsion component parameters, and the attitude execution unit reachable range. The curve segments that exceed the execution reachable range are removed, and buffer segments are inserted in the load fluctuation area to obtain the smooth switching trajectory and the power scheduling trajectory: the power upper limit , the power lower limit , the power change rate upper limit , the phase reachable domain of the attitude execution unit , and the torque margin constitute the reachable domain. Projection and clipping are performed on the curve prototype, and the load fluctuation area is time-expanded and slope-degraded, formula: ; ; ;

[0078] wherein, is an interval projection operator, is a rate clipping operator; is a phase reachable range of the attitude execution unit, is a set of allowable torques; is a load fluctuation zone indicator, taking 1 when in the load fluctuation zone; is a buffer section widening and slope degradation coefficient; is an effective lift power duration after widening, is a degraded effective rate upper limit. After the above semantic alignment, reachable domain projection, and fluctuation zone buffering, together constitute the smooth switching trajectory and power scheduling trajectory, which can realize the continuous balanced transition of the powerboat lift power and the unmanned ship load reduction under the conditions of meeting the switching preparation order, the propulsion component constraints, and the attitude execution unit reachable range throughout the whole process.

[0079] S160, collect trajectory deviation, attitude deviation, contact load, and execution feedback throughout the whole interaction process to construct a causal graph, and complete the interactive replacement of the unmanned ship powerboat based on the smooth switching trajectory, the power scheduling trajectory, and the causal graph.

[0080] Specifically, the whole interaction process refers to the whole time sequence interval from the discovery of the energy powerboat by the unmanned ship, approach, adhesion, locking, power switching to decoupling steady state, covering the data and instruction flow closed loop of perception, prediction, planning, execution, and rollback. For example, during the nearshore surge enhancement, each rolling sampling period from the approach stage to the completion of the locking is included in the monitoring and recording of the whole interaction process. The causal graph refers to a graph structure organized by trajectory deviation nodes, attitude deviation nodes, contact load nodes, execution feedback nodes, and sea state disturbance nodes in a time-ordered causal connection, used to explain the source of deviation, verify the control effect, and trigger the rollback strategy. For example, the sea state disturbance node points to the attitude deviation node, which in turn points to the execution feedback node, indicating that the attitude mismatch is caused by external surge and has been locally compensated.

[0081] Further, continuous observation data is collected by the trajectory deviation collector, the attitude deviation collector, the contact load collector, and the execution feedback collector under the unified timestamp constraint, and the trajectory deviation sequence, the attitude deviation sequence, the contact load sequence, and the execution feedback sequence are generated on the unified sampling grid; specifically, after aligning the relative position observation sequence and the reference trajectory on the unified time grid, the trajectory deviation vector sequence is obtained, denoted as After aligning the yaw, pitch, and yaw angle observations with the reference attitude, the attitude deviation vector sequence is obtained, denoted as... The contact loads measured by the contact array are integrated into a contact load sequence, denoted as . The local execution status is encoded into an execution feedback sequence, denoted as... ,in The above four types of sequences are organized with the same observation window length and the same sampling period to ensure consistency of the time base across channels. The trajectory deviation and attitude deviation are calculated using the following formula and used as direct input for the second step of event-based annotation: ; ;

[0082] In the formula, For the first The relative position observation vector at each sampling time; For the first The reference relative position vector at each sampling time; These are the observed and reference quantities for the yaw angle, pitch angle, and yaw angle, respectively; For the first Contact load vector at each sampling time; For the first The execution feedback status code at each sampling time.

[0083] Based on the four types of sequences output from the first step, event-based annotation is performed to construct an initial graph. Within a sliding window, standardized deviations and rates of change are calculated for trajectory and attitude deviations, and impact and hysteresis indicators are calculated for contact loads and execution feedback. This identifies approach deviation events, attitude mismatch events, contact impact events, and execution hysteresis events, and generates an initial graph containing trajectory deviation nodes, attitude deviation nodes, contact load nodes, and execution feedback nodes. The discriminant and triggering criteria are defined as follows, with event indicators serving as the basis for generating edges in the initial graph: ; ; ; ;

[0084] In the formula, The mean and standard deviation of the trajectory deviation norm within the window; The mean and standard deviation of the attitude deviation norm within the window; Standardized threshold; The threshold for the rate of change of deviation; The threshold for contact load variation; respectively, are the event indicators of approaching deviation, attitude mismatch, contact impact and execution lag. The directed edges are established between the four types of nodes with the constraints of event indicators and time sequence, and the initial graph is obtained as the input of the third step of graph alignment and convergence check.

[0085] The smooth switching trajectory and the power scheduling trajectory are graphically aligned with the initial graph obtained in the second step to form a causal graph under the constraint of contact load, and the trajectory deviation nodes, attitude deviation nodes and execution feedback nodes are checked for convergence within the rolling sampling period. When aligning, the optimal time offset of the reference trajectory and the observation sequence is first estimated, then the allowable interval of the contact load is used as the boundary condition to filter invalid edges, and then the convergence criterion is calculated and the completion state of the unmanned ship power boat interactive replacement is output. The time alignment, load constraint and convergence criterion are as follows: ; ; ; ;

[0086] In the formula, is the optimal time offset, which is used to align the smooth switching trajectory and the power scheduling trajectory with the observation on the time axis; is the lower limit and upper limit of the contact load component; is the starting index of the convergence evaluation window; is the window length; are the average norms of the trajectory deviation and attitude deviation in the window, respectively; is the proportion of "completion" of execution feedback in the window; is the deviation convergence threshold; is the execution success rate threshold; is the interactive replacement completion state indicator. Through the above chain implementation, the four types of sequence inputs of the first step drive the eventization and initial graph construction of the second step, the initial graph of the second step and the smooth switching trajectory and the power scheduling trajectory jointly drive the graph alignment and convergence determination of the third step, and finally form a verifiable completion state output.

[0087] The application also provides an unmanned ship power boat interactive replacement device, which is described with reference to Figure 2 , Figure 2A module schematic diagram of an unmanned ship power boat interactive replacement device is provided for an embodiment of the present application. The device is a server, and the server includes an acquisition module 21 and a processing module 22. The acquisition module 21 is configured to acquire multi-modal observation streams of an unmanned ship, a power boat, and an interactive environment, and extract a candidate factor set from the multi-modal observation streams based on a disturbance field reconstruction pipeline. The processing module 22 is configured to input the candidate factor set into a short-time trajectory prediction and attitude prediction engine, generate trajectory prediction and attitude prediction results in combination with sea state disturbance information, broadcast the trajectory prediction and attitude prediction results to the unmanned ship and the power boat under a multi-modal interaction protocol, and generate a proximity control instruction set according to an interactive intention sharing stream fed back by the unmanned ship and the power boat after the broadcasting. The processing module 22 is further configured to issue the proximity control instruction set to the unmanned ship and the power boat, correct the trajectory prediction and attitude prediction results under the support of disturbance correction services provided by a buoy node, output a disturbance-corrected proximity state stream, and trigger a docking geometric guidance process when a proximity error falls within a threshold range to complete a locking verification. The processing module 22 is further configured to, after the docking locking state is confirmed, adjust the power output of the power boat and simultaneously reduce the power output of the unmanned ship according to a continuous power distribution curve based on a contact load spectrum and a locking state, to obtain a smooth switching trajectory and a power scheduling trajectory. The processing module 22 is further configured to collect trajectory deviation, attitude deviation, contact load, and execution feedback throughout the entire interactive process, construct a causal graph, and complete the unmanned ship power boat interactive replacement based on the smooth switching trajectory, the power scheduling trajectory, and the causal graph.

[0088] It should be noted that the device provided in the above embodiments, when realizing its functions, is only exemplified by the above division of functional modules. In actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above-described functions. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be described here.

[0089] The present application also provides an electronic device, referring to Figure 3 , Figure 3 A structural schematic diagram of an electronic device is provided for an embodiment of the present application. The electronic device can include at least one processor 31, at least one network interface 34, a user interface 33, a memory 35, and at least one communication bus 32.

[0090] The communication bus 32 is configured to realize the connection and communication between the components.

[0091] The user interface 33 can include a display, a camera, and optionally a standard wired interface and a wireless interface.

[0092] The network interface 34 can optionally include a standard wired interface and a wireless interface (e.g., a Wi-Fi interface).

[0093] The processor 31 can include one or more processing cores. The processor 31 connects various parts of the server through various interfaces and lines, and performs various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 35, and calling data stored in the memory 35. Optionally, the processor 31 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 31 can be integrated with a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU is mainly used to process an operating system, a user interface, and an application program. The GPU is used to render and draw content to be displayed on the display. The modem is used to process wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 31, but can be implemented by a separate chip.

[0094] The memory 35 can include a random access memory (RAM) and a read-only memory (ROM). Optionally, the memory 35 includes a non-transitory computer-readable storage medium. The memory 35 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 35 can include a program storage area and a data storage area. The program storage area can store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc. The data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 35 can optionally be at least one storage device located away from the above-mentioned processor 31. For example, Figure 3As shown, the memory 35, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for an interactive replacement method for an unmanned powered boat.

[0095] exist Figure 3 In the electronic device shown, the user interface 33 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 31 can be used to call the application program stored in the memory 35 for an interactive replacement method of an unmanned boat powered boat. When executed by one or more processors, the electronic device executes one or more methods as described in the above embodiments.

[0096] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0097] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for interactively replacing unmanned surface vessel (USV) powered boats, characterized in that, The method includes: Acquire multimodal observation streams from unmanned vessels, powered boats, and interactive environments, and extract candidate factor sets from the multimodal observation streams based on the perturbation field reconstruction pipeline; The candidate factor set is input into the short-time trajectory prediction and attitude prediction engine, and the trajectory prediction and attitude prediction results are generated by combining the sea state disturbance information. The trajectory prediction and attitude prediction results are broadcast to the unmanned vessel and the powered boat under a multimodal interaction protocol, and a set of proximity control commands is generated based on the interactive intent sharing stream fed back by the unmanned vessel and the powered boat after the broadcast. The approach control command set is sent to the unmanned vessel and the powered boat. With the support of the disturbance correction service provided by the buoy node, the trajectory prediction and attitude prediction results are corrected, the disturbance-corrected approach state flow is output, and the docking geometry guidance process is triggered when the approach error falls within the threshold range to complete the lock-in verification. After confirming the docking lock status, based on the contact load spectrum and the lock status, the power output of the power boat is increased and the power output of the unmanned vessel is simultaneously reduced according to the continuous power distribution curve, so as to obtain a smooth switching trajectory and power scheduling trajectory. Throughout the interaction process, trajectory deviation, attitude deviation, contact load, and execution feedback are collected to construct a causal graph. Based on the smooth switching trajectory, the power scheduling trajectory, and the causal graph, the interactive switching of the unmanned vessel's powered boat is completed.

2. The interactive replacement method for unmanned surface vessel powered boats according to claim 1, characterized in that, The acquisition of multimodal observation streams from unmanned surface vessels, powered boats, and the interactive environment, and the extraction of candidate factor sets from these multimodal observation streams based on a perturbation field reconstruction pipeline, specifically includes: Position, velocity, acceleration, attitude, sea state disturbance, and structural response information are collected by visual sensors, acoustic positioning sensors, inertial navigation units, satellite and short baseline joint positioning components, wave spectrum measurement units, and propulsion component strain gauges and temperature sensors. All information is then uniformly clock-aligned, time-domain interpolated, and missing data is filled on a distributed soft bus to form a multimodal observation stream under consistent sampling constraints. Cross-domain alignment and noise suppression are performed on the multimodal observation stream to generate a cleaned multimodal observation stream enhancement result; Based on the enhanced results of the multimodal observation flow, event-based processing is performed to label track change events, attitude change events, surge impact events, roll amplification events, and propulsion load surge events, thereby obtaining an intermediate cue set segmented by short-time windows; The intermediate clue set is input into the disturbance field reconstruction pipeline, and scene tags and far-field wind and wave forecast information from local sea state observations, nearby powered boat broadcasts, and historical experience caches are integrated to generate a disturbance element sequence that includes surges, surge superpositions, and local vortices. Based on the perturbation element sequence, the channel features in the multimodal observation flow enhancement results are constrained and aggregated. Attitude coupling factors related to roll, pitch, and yaw are extracted, and propulsion coupling factors related to thrust amplitude, thrust direction, and thrust pulsation are extracted. Robustness checks and redundancy compression are performed on the attitude coupling factors and the propulsion coupling factors to obtain the candidate factor set.

3. The interactive replacement method for unmanned surface vessel powered boats according to claim 2, characterized in that, The step of inputting the candidate factor set into the short-time trajectory prediction and attitude prediction engine, and combining it with sea state disturbance information to generate trajectory prediction and attitude prediction results, specifically includes: Missing test completion, anomalous mutation smoothing and long-term drift correction are performed on the candidate factor set to form a feature sequence package, and the feature sequence package is input into the short-term trajectory prediction and attitude prediction engine to establish the context of the current rolling prediction cycle; Based on the disturbance element sequence and the context, a multi-scenario candidate trajectory cluster and a multi-scenario candidate attitude cluster are generated. Combined with the constraint set composed of geometric shape constraints, propulsion component constraints, docking guide constraints and buffer component constraints, candidates that do not meet the geometric alignment, velocity range, angular velocity range and load threshold are eliminated, and a candidate set that meets the execution reachability is output. Uncertainty assessment and stage priority ranking are performed on the candidate set, and the main trajectory and main attitude are selected as the trajectory prediction and attitude prediction results.

4. The interactive replacement method for unmanned surface vessel powered boats according to claim 1, characterized in that, The step of broadcasting the trajectory prediction and attitude prediction results to the unmanned vessel and the powered boat under a multimodal interaction protocol, and generating a proximity control command set based on the interactive intent sharing stream fed back by the unmanned vessel and the powered boat after broadcasting, specifically includes: The trajectory prediction and attitude prediction results are encapsulated into a data packet containing a unified timestamp, a rolling prediction cycle identifier, a trajectory segment sequence, an attitude segment sequence, a disturbance scenario label, a confidence label, and an execution time label, and the data packet is sent to the unmanned vessel and the powered boat in a deterministic and ordered broadcast manner on a distributed soft bus. After the unmanned vessel and the powered boat receive the data packet, they obtain an interactive intent-sharing stream preview frame that includes trajectory weighting, attitude weighting, propulsion redundancy, attitude adjustment margin, energy margin, and health status labels. The interactive intent sharing stream preview frames are aggregated and semantic consistency checks and conflict alignments are performed to obtain a conflict summary containing directional conflicts, timing conflicts and resource conflicts. Based on the conflict summary, supplementary attitude target priority and propulsion component temperature rise limit constraints are requested from the unmanned vessel and the powered boat to generate the interactive intent sharing stream. The interactive intent sharing stream is superimposed with the trajectory prediction and attitude prediction results, sea state disturbance scenario labels, unmanned surface vessel constraints, and powered boat constraints to form a constraint sequence sorted by priority. The approach control command set is generated based on the constraint sequence. The approach control command set includes velocity constraint command, yaw constraint command, pitch constraint command, approach corridor correction command, attitude target zone correction command, and safety buffer setting command.

5. The interactive replacement method for unmanned surface vessel powered boats according to claim 1, characterized in that, The process involves sending the proximity control command set to the unmanned vessel and the powered boat, correcting the trajectory and attitude prediction results with the support of the disturbance correction service provided by the buoy node, outputting the disturbance-corrected proximity state flow, and triggering the docking geometry guidance process when the proximity error falls within a threshold range to complete the lock-on verification. Specifically, this includes: The proximity control command set is sent to the execution interfaces corresponding to the unmanned vessel and the powered boat. The execution interface completes local atomic switching based on the unified effective time and execution time tag, and generates execution receipt and safety verification result. Through the disturbance correction service of the buoy node, based on the execution receipt and safety verification results, and combined with the near-field sea state observation channel and historical disturbance cache, a set of correction fragments is generated. At the same time, the trajectory fragment sequence and attitude fragment sequence are aligned with errors to obtain the corrected trajectory fragments and attitude fragments. The corrected trajectory segment and attitude segment, the navigation state, attitude state and propulsion state of the unmanned vessel and the powered boat are fused together to output a disturbance-corrected approach state flow. When the approach state flow meets the threshold conditions of geometric alignment state, velocity boundary and angular velocity boundary, the docking geometry guidance process is triggered. In the docking geometry guidance process, docking guidance instructions and buffer configuration plans are generated based on the proximity state flow, and the docking guidance instructions and buffer configuration plans are sent to the guidance execution units of the unmanned vessel and the powered boat to perform low-speed fitting, lateral fine-tuning and longitudinal fine-tuning. During the low-speed bonding, lateral fine-tuning and longitudinal fine-tuning processes, if the contact detection subprocess confirms that the contact point distribution and contact consistency meet the preset conditions, then the locking verification is performed, and the contact load spectrum and the locking state are output.

6. The interactive replacement method for unmanned surface vessel powered boats according to claim 1, characterized in that, After confirming the docking lock-in state, based on the contact load spectrum and the lock-in state, the power output of the powered boat is increased according to the continuous power distribution curve, while the power output of the unmanned vessel is simultaneously reduced, resulting in a smooth switching trajectory and a power scheduling trajectory. Specifically, this includes: The contact load spectrum is analyzed into a lock-up stability region, a buffer attenuation region, and a load fluctuation region, and a switching preparation order carrying the upper limit of the thrust change rate and the upper limit of the torque change rate is generated. Under the constraints of the switching preparation order, the rising curve of the power output of the powered boat and the falling curve of the power output of the unmanned vessel are instantiated, and a phase reference trajectory and a torque balance reference trajectory are generated to form a set of curve prototypes. The curve prototype set is semantically aligned with the propulsion component parameters of the powered boat, the propulsion component parameters of the unmanned vessel, and the reachable range of the attitude execution unit. Curve segments that exceed the reachable range are removed, and buffer segments are inserted in the load fluctuation area to obtain the smooth switching trajectory and the power scheduling trajectory.

7. The interactive replacement method for unmanned surface vessel powered boats according to claim 1, characterized in that, The process of collecting trajectory deviation, attitude deviation, contact load, and execution feedback throughout the interaction to construct a causal graph, and then completing the interactive switching of the unmanned surface vessel's powered boat based on the smooth switching trajectory, the power scheduling trajectory, and the causal graph, specifically includes: Continuous observation data is collected under a unified timestamp constraint using a trajectory deviation collector, attitude deviation collector, contact load collector, and execution feedback collector. Based on the continuous observation data, a trajectory deviation sequence, attitude deviation sequence, contact load sequence, and execution feedback sequence are generated. The trajectory deviation sequence, attitude deviation sequence, contact load sequence, and execution feedback sequence are annotated with events to identify approach deviation events, attitude mismatch events, contact impact events, and execution lag events, so as to construct an initial map containing trajectory deviation nodes, attitude deviation nodes, contact load nodes, and execution feedback nodes. The smooth switching trajectory, the power scheduling trajectory, and the initial map are graph aligned to generate a causal map under contact load constraints. The convergence of the trajectory deviation node, the attitude deviation node, and the execution feedback node is evaluated in the rolling sampling period to determine the completion status of the interactive replacement of the unmanned vessel powered boat based on the convergence.

8. An interactive device for changing the power of an unmanned surface vessel (USV), characterized in that, The device is used to perform the interactive replacement method for unmanned surface vessels powered by a motor as described in any one of claims 1 to 7, the device comprising an acquisition module (21) and a processing module (22), wherein, The acquisition module (21) is used to acquire multimodal observation streams of unmanned boats, powered boats and interactive environments, and extract candidate factor sets from the multimodal observation streams based on the perturbation field reconstruction pipeline; The processing module (22) is used to input the candidate factor set into the short-term trajectory prediction and attitude prediction engine, and generate trajectory prediction and attitude prediction results by combining the sea state disturbance information. The processing module (22) is also used to broadcast the trajectory prediction and attitude prediction results to the unmanned vessel and the powered boat under the multimodal interaction protocol, and generate a set of proximity control instructions based on the interactive intent sharing stream fed back by the unmanned vessel and the powered boat after the broadcast. The processing module (22) is also used to send the proximity control command set to the unmanned vessel and the powered boat, correct the trajectory prediction and attitude prediction results with the support of the disturbance correction service provided by the buoy node, output the disturbance-corrected proximity state flow, and trigger the docking geometry guidance process when the proximity error falls within the threshold range to complete the lock-in verification. The processing module (22) is also used to, after confirming the docking lock state, adjust the power output of the power boat according to the contact load spectrum and the lock state, and simultaneously reduce the power output of the unmanned vessel according to the continuous power distribution curve, so as to obtain a smooth switching trajectory and a power scheduling trajectory. The processing module (22) is also used to collect trajectory deviation, attitude deviation, contact load and execution feedback during the entire interaction process to construct a causal graph, and to complete the interactive replacement of the unmanned boat power boat based on the smooth switching trajectory, the power scheduling trajectory and the causal graph.

9. An electronic device, characterized in that, The electronic device includes a processor (31), a memory (35), a user interface (33), and a network interface (34). The memory (35) is used to store instructions. The user interface (33) and the network interface (34) are both used to communicate with other devices. The processor (31) is used to execute the instructions stored in the memory (35) to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 7.

Images