Underwater wet-mate connector autonomous docking method and system

Abstract

The application discloses an underwater wet plug connector autonomous docking method and system, comprising the following steps: obtaining the structure model, material parameters, environmental parameters, execution constraints and target indicators of the target underwater wet plug connector; constructing parameterized digital objects containing plug-in, guidance, sealing and other parameters based on the structure model; constructing a simulation scene containing approach, alignment, insertion, sealing penetration, abnormal back-off retry, locking scene; setting a large model multi-agent collaboration module, decomposing the task and performing structure understanding, physical modeling and other work; executing simulation to obtain a virtual state quantity sequence and generate a candidate docking strategy; criticizing and checking the candidate strategy, if the requirement is not met, optimizing the parameters and re-simulating, if the requirement is met, outputting the optimal docking strategy, optimization parameters and applicable working conditions and completing autonomous docking. The application realizes docking strategy generation full-process automation, reduces manual dependence, fills the virtual-real gap, and improves research and development efficiency and engineering reliability.

Description

Technical Field

[0001] This invention relates to the field of underwater connector technology, and more specifically, to an autonomous docking method and system for underwater wet-plug connectors. Background Technology

[0002] Current autonomous docking technology for underwater wet-plug connectors faces three core problems. First, it lacks an integrated automated technology system. The processes of simulation construction, docking strategy generation, and multi-objective joint optimization are fragmented, making it impossible to automate the entire process and requiring manual intervention at each step. Second, strategy design relies excessively on human experience. Engineers need to repeatedly debug parameters and iterate solutions, resulting in long design cycles, low efficiency, and the quality of strategies being heavily influenced by individual experience levels, making it difficult to form standardized and reusable design capabilities. Third, there is a serious "virtual-real gap." Existing simulations cannot accurately reproduce complex underwater environments and actual engineering conditions. Many strategies that converge in simulations fail in field applications, and there is a lack of automatic identification mechanisms for such invalid results, leading to wasted R&D resources and increased risks in engineering applications. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide an autonomous docking method and system for underwater wet-plug connectors. By constructing parameterized digital objects and a multi-scenario simulation system, and combining large-scale models with multi-agent collaboration, the invention achieves automated docking strategy generation and optimization throughout the entire process, thus solving the problems of fragmented processes, reliance on human experience, and mismatch between virtual and real technologies in existing technologies.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for autonomous docking of an underwater wet-plug connector includes the following steps: Step 1: Obtain the structural model, material parameters, environmental parameters, execution constraints, and target indicators of the target underwater wet-plug connector; Step 2: Construct a connector parameterized digital object based on the structural model from Step 1. The parameterized digital object includes at least one or more of the following: mating structure parameters, guiding structure parameters, sealing structure parameters, and locking structure parameters. Step 3: Construct a simulation scenario for the autonomous docking process of the connector. The simulation scenario shall include at least one or more of the following scenarios: approach scenario, alignment scenario, insertion scenario, seal penetration scenario, abnormal retreat and retry scenario, and locking scenario. Step 4: Set up a large-scale multi-agent collaborative module to decompose the autonomous docking task and execute structural understanding, physical modeling, simulation execution, result critique, strategy generation, and parameter optimization respectively. Step 5: Perform an autonomous docking process simulation to obtain the corresponding virtual state variable sequence; Step 6: Generate candidate docking strategies based on the virtual state variable sequence; Step 7: Critically evaluate and verify the candidate docking strategies, and determine whether they meet the preset constraints and target indicators. When the candidate docking strategy does not meet the preset constraints and / or target indicators, optimize one or more of the action parameters, structural parameters, and operating condition parameters, and perform resimulation and iteration. When the candidate docking strategy meets the preset constraints and / or target indicators, output the optimal docking strategy, optimized parameters, and applicable operating conditions that meet the target requirements. Based on the output optimal docking strategy, optimized parameters, and applicable operating conditions, complete the autonomous docking of the underwater wet-plug connector.

[0005] Another aspect of the present invention provides an autonomous docking system for underwater wet-plug connectors, equipped with the aforementioned autonomous docking method for underwater wet-plug connectors, comprising: The task input module is used to execute step one to receive input information related to the autonomous docking task; The parametric modeling module is used to perform step two, converting the input connector model into a parametric digital object that can be used for simulation and optimization. The scenario construction module is used to execute step three to construct different simulation scenarios of the docking process based on the autonomous docking task definition. The large-scale model multi-agent collaboration module is used to execute step four to organize multiple functional agents to decompose, reason, model, critique, and optimize the task; The simulation execution module is used to execute step five, which simulates the autonomous docking process based on the parameterized digital objects and the scene construction results. The critique and verification module is used to perform step six to judge the engineering and physical rationality of the simulation results and candidate strategies; The strategy generation module is used to execute step seven to generate candidate docking strategies and the optimal docking strategy based on the task objective, scenario information and simulation results. The parameter optimization module is used to execute step seven to optimize one or more of the motion parameters, structural parameters, and operating parameters according to a preset objective function. The knowledge base module is used to perform step seven to store successful strategies, failure cases, parameter ranges, applicable scenarios, and migration rules. The output results module is used to execute step seven to output the optimal docking strategy, optimization parameters, and applicable operating conditions.

[0006] As a further improvement to the system, the large-scale model multi-agent collaborative module includes: The Collaboration Manager agent is used to dynamically allocate task order and invocation relationships based on the current task context. A task planning agent is used to break down the input autonomous docking task into several sub-tasks; A structural understanding agent is used to perform semantic parsing of the geometry and assembly relationships of connectors; A physical modeling agent is used to construct a physical model that adapts to the autonomous docking process based on the results of structural analysis. The simulation execution agent is used to call the simulation execution module to solve the problem based on the simulation task description generated by the physical modeling agent, and then return the simulation results. The result critique agent is used to critique and verify the simulation results; The parameter optimization agent is used to optimize relevant parameters based on the criticism results and target metrics given by the result criticism agent. The knowledge-accumulating intelligent agent is used to write successful strategies, failure cases, corresponding parameter ranges, and migration rules into the knowledge base module.

[0007] As a further improvement to the system, the parametric modeling module transforms the input connector model into a parametric digital object that can be used for simulation and optimization. Specifically: The mating parameter set, including pin diameter, socket diameter, and mating stroke, is used to determine the contact relationship and mating resistance characteristics during the mating process; The set of guidance parameters, including radial alignment deviation, angular alignment deviation, and rotational alignment deviation, is used to determine the capture capability and alignment tolerance during the approach and alignment phases. A set of sealing parameters, including sealing resistance, seal thickness, and insertion / extraction speed, is used to determine the resistance, deformation, and recoverability during the seal penetration phase. The locking parameter set, including locking stroke, docking force, and disconnecting force, is used to determine the triggering conditions and locking reliability of the locking action.

[0008] As a further improvement to the system, the scenario construction module builds a simulation scenario for the autonomous docking process, including the following stages: During the approach phase, the end of the actuator carrying the plug gradually approaches the socket end, but contact has not yet occurred; During the alignment phase, the plug and socket ends are spatially aligned using attitude correction actions. During the insertion phase, the plug end begins to enter the socket end, initial contact occurs, and insertion force is generated. During the sealing penetration stage, the plug end penetrates the sealing area, causing local deformation of the sealing structure. During the abnormal rollback and retry phase, if the strategy execution process encounters issues such as jamming, excessive insertion force, locking failure, or alignment failure, the system will perform a rollback action, realign, and then try again.

[0009] During the locking phase, the connector completes the preset insertion depth and triggers the locking action; Beneficial effects: This invention, by constructing a fully integrated autonomous docking technology system, deeply integrates simulation construction, strategy generation, and parameter optimization, achieving an automated closed loop from task input to optimal strategy output, eliminating the need for manual intervention and significantly improving R&D efficiency. It replaces the traditional experience-driven design mode with a large-scale model multi-agent collaborative mechanism, completing structural analysis, physical modeling, simulation execution, and parameter optimization through multi-agent collaboration, significantly shortening the design cycle, eliminating the impact of individual experience differences on strategy quality, and forming standardized and reusable design capabilities. Simultaneously, through multi-scenario full-process simulation and critical verification mechanisms, it accurately reproduces complex underwater working conditions, automatically identifies and corrects ineffective strategies, effectively bridging the "virtual-real gap," reducing R&D resource waste, and lowering engineering application risks.

[0010] The autonomous docking system provided by this invention achieves decoupling and collaboration of various functional units through modular design, realizes digital abstraction of connector structure through parametric modeling module, ensures efficient decomposition and execution of tasks through large model multi-agent collaboration module, and realizes the accumulation and transfer of experience through knowledge base module, further improving the system's versatility and scalability. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the overall architecture of the platform of the present invention; Figure 2 This is a schematic diagram of multi-agent role collaboration in the large-scale model of the present invention; Figure 3 This is a schematic diagram of the parameterized digital object of the underwater wet-plug connector of the present invention; Figure 4 This is a schematic diagram of a simulation scenario of the autonomous docking process of the present invention; Figure 5 This is a flowchart of the simulation-critique-strategy generation-optimization closed loop of the present invention; Figure 6 This is a schematic diagram illustrating the strategy migration and knowledge base update of the present invention. Detailed Implementation

[0012] The present invention will now be described in further detail with reference to the embodiments shown in the accompanying drawings.

[0013] Figure 1 This is a schematic diagram of the overall architecture of the platform of this invention. The diagram shows the connection relationships and data flow between the task input module, parametric modeling module, scene construction module, large model multi-agent collaboration module, simulation execution module, criticism and verification module, strategy generation module, parameter optimization module, knowledge base module, and output results.

[0014] Figure 2This is a schematic diagram illustrating the multi-agent role collaboration in the large-scale model of this invention. The diagram shows the collaborative relationships between the collaboration manager agent, task planning agent, structural understanding agent, physical modeling agent, simulation execution agent, result criticism agent, parameter optimization agent, and knowledge accumulation agent.

[0015] Figure 3 This is a schematic diagram of the parameterized digital object of the underwater wet-plug connector of the present invention. The figure shows the mating structure, guiding structure, sealing structure and locking structure of the underwater wet-plug connector, as well as the corresponding mating parameter set, guiding parameter set, sealing parameter set and locking parameter set, and illustrates the dual-stage oil filling pressure balance, penetrable self-sealing and automatic interlocking docking feature structure.

[0016] Figure 4 This is a schematic diagram of a simulation scenario of the autonomous docking process of the present invention. The diagram shows the approach stage, alignment stage, insertion stage, seal penetration stage, abnormal retreat and retry stage, and locking stage, and illustrates virtual state quantities such as relative pose, contact state, insertion force, seal deformation, and locking state.

[0017] Figure 5 This is a flowchart illustrating the closed-loop process of simulation-critique-policy generation-optimization for this invention. The flowchart shows the steps including task definition, parametric modeling, scenario construction, process simulation, result criticism, candidate policy generation, parameter optimization, resimulation and iteration, parameter correction, and outputting the optimal policy, as well as the constraint satisfaction and policy availability judgment nodes.

[0018] Figure 6 This diagram illustrates the strategy migration and knowledge base update process of this invention. It shows the successful strategy set and failure case set of connector model A being written into the knowledge base, and the strategy templates and initial parameter suggestions being output to connector models B and C through feature extraction and similarity evaluation, as well as the update and backfeeding process.

[0019] Based on the above figures, this embodiment provides a platform and method for simulating, generating, and optimizing the autonomous docking process of underwater wet-plug connectors based on a large-model, multi-agent collaborative approach. This platform addresses the challenges of autonomous docking tasks for underwater wet-plug connectors, including complex modeling of the docking process, reliance on human experience for strategy design, disconnect between simulation and optimization, and a lack of automatic criticism and correction mechanisms for result evaluation. It proposes a process simulation, strategy generation, and parameter optimization scheme involving multiple functional agents built on a large model collaboratively.

[0020] In this invention, the "large model multi-agent" is preferably a group of functional agent units constructed based on a large language model, a multimodal large model, or other generative large models. These agent units, through role division and information interaction, jointly complete tasks such as task planning, structural understanding, physical modeling, simulation execution, result critique, parameter optimization, and knowledge accumulation. It should be noted that the "large model multi-agent" in this invention differs from the multiple physical actuators or control nodes in traditional distributed control; rather, it is a cognitive collaborative unit oriented towards autonomous docking with connectors for simulation and strategy generation tasks.

[0021] 1. Platform Overall Structure like Figure 1 As shown, the platform of the present invention mainly includes: a task input module 101, a parametric modeling module 102, a scene construction module 103, a large model multi-agent collaboration module 104, a simulation execution module 105, a criticism and verification module 106, a strategy generation module 107, a parameter optimization module 108, a knowledge base module 109, and an output result module 110.

[0022] The task input module 101 is used to receive input information related to the autonomous docking task. Preferably, the input information includes, but is not limited to: connector model, material parameters, environmental parameters, execution constraints, and target indicators. The connector model can be a three-dimensional geometric model, a two-dimensional cross-sectional model, an assembly relationship model, or a combination thereof; the material parameters include elastic modulus, Poisson's ratio, sealing material parameters, friction parameters, contact parameters, etc.; the environmental parameters include water depth, external pressure, flow velocity disturbance, temperature, etc.; the execution constraints include the upper limit of actuator speed, attitude correction range, maximum allowable insertion force, etc.; the target indicators include docking success rate, docking time, peak insertion force, lock completion rate, and robustness, etc.

[0023] The parametric modeling module 102 is used to convert the input connector model into a parametric digital object that can be used for simulation and optimization. Preferably, the parametric modeling module 102 can decompose the connector into several functional substructures and assign structural parameters to each. In this embodiment, the functional substructures include at least a mating structure, a guiding structure, a sealing structure, and a locking structure.

[0024] The scenario construction module 103 is used to construct different simulation scenarios of the docking process according to the definition of the autonomous docking task. Preferably, the scenario construction module 103 can generate one or more scenarios such as ideal alignment scenario, eccentricity scenario, angular deviation scenario, disturbance scenario, locking failure scenario, and backtracking retry scenario.

[0025] The large-scale model multi-agent collaboration module 104 is the core module of this invention, used to organize multiple functional agents to decompose, reason, model, critique, and optimize tasks. In a preferred embodiment, the large-scale model multi-agent collaboration module 104 includes at least a task planning agent, a structural understanding agent, a physical modeling agent, a simulation execution agent, a result critique agent, a parameter optimization agent, a knowledge accumulation agent, and a collaboration manager agent.

[0026] The simulation execution module 105 is used to perform autonomous docking process simulation based on the parameterized digital object and scene construction results. The simulation execution module 105 can call at least one solver, including but not limited to rigid body or rigid-flexible body kinematics solver, contact collision simulation solver, structural mechanics finite element solver, local sealing deformation solver, reduced-order approximation solver, and multiphysics co-simulation solver.

[0027] The critique and verification module 106 is used to judge the engineering and physical rationality of simulation results and candidate strategies. Unlike conventional processes that rely solely on convergence judgment, the critique and verification module 106 in this invention can identify invalid success results where the simulation has been completed but are not engineering-usable.

[0028] The strategy generation module 107 is used to generate candidate docking strategies and the optimal docking strategy based on the task objective, scenario information, and simulation results. The candidate docking strategies preferably include one or more of the following: approach action, attitude correction action, insertion action, micro-rotation action, retreat action, retry action, and locking action.

[0029] The parameter optimization module 108 is used to optimize one or more of the action parameters, structural parameters, and operating condition parameters according to a preset objective function. The knowledge base module 109 is used to store successful strategies, failure cases, parameter ranges, applicable scenarios, and migration rules. The output result module 110 is used to output the optimal docking strategy, optimized parameters, and applicable operating conditions.

[0030] 2. Multi-agent collaborative mechanism in large-scale models like Figure 2 As shown, in a preferred embodiment, the large model multi-agent collaboration module 104 coordinates multiple functional agents to work collaboratively through the collaboration manager agent 201.

[0031] The collaborative manager agent 201 is used to dynamically allocate task order and invocation relationships based on the current task context. Preferably, the collaborative manager agent 201 can dynamically schedule among agents based on task complexity, simulation status, criticism results, and optimization stage.

[0032] The task planning agent 202 is used to break down the input autonomous docking task into several sub-tasks. For example, for a given underwater wet-plug connector autonomous docking task, the task planning agent 202 can break it down into a connector structure semantic parsing sub-task, a scenario construction sub-task, a contact plugging simulation sub-task, a seal penetration simulation sub-task, a locking action simulation sub-task, a result criticism sub-task, a parameter optimization sub-task, and a knowledge accumulation sub-task.

[0033] The structural understanding agent 203 is used to perform semantic parsing of the connector's geometry and assembly relationships. Preferably, the structural understanding agent 203 identifies one or more of the following structures: guide segment, pin or socket area, sealing ring or sealing lip, locking ramp or interlocking component, pressure balance related cavity or compensation structure. In one embodiment, the structural understanding agent 203 extracts key parameters such as guide cone angle, mating tolerance, sealing segment geometry, and locking stroke from the connector model and outputs them to the physical modeling agent 204.

[0034] The physical modeling agent 204 is used to construct a physical model adapted to the autonomous docking process based on the structural analysis results. Preferably, the physical modeling agent 204 establishes different sub-models according to different stages, including a kinematic model for the approach and alignment stage, a contact collision model for the insertion stage, a local deformation model for the sealing and penetration stage, a mechanism motion model for the locking stage, and a robustness analysis model under external disturbances. The physical modeling agent 204 can also automatically adjust the model complexity according to task requirements, for example, calling a reduced-order model during rapid screening and a finite element model during high-precision analysis.

[0035] The simulation execution agent 205 is used to call the simulation execution module 105 to perform the solution based on the simulation task description generated by the physical modeling agent 204, and then return the simulation results. The simulation execution agent 205 can output one or more of the following virtual state quantities: relative pose, contact start time, contact area distribution, insertion force curve, sealing deformation, locking state quantity, local stress or strain, whether docking is completed, and failure reason label.

[0036] The result criticism agent 206 is used to critique and verify the simulation results. The result criticism is not limited to whether the solution converges, but further determines whether the generated results meet engineering application requirements. For example, if the simulation results meet numerical convergence but exhibit one of the following: excessively high peak insertion force, significant local interference, excessive deformation of the sealing structure, locking not completed within the effective stroke, strategy failure under minor perturbations, or excessively narrow tolerance window, the result criticism agent 206 can still determine that the strategy is invalid.

[0037] The parameter optimization agent 207 is used to optimize relevant parameters based on the criticism results and target indicators given by the result criticism agent 206. In a preferred embodiment, the parameter optimization agent 207 can jointly optimize one or more of the following: approach speed, insertion speed, micro-rotation angle, back-off distance, lock trigger threshold, number of retries, guidance parameters, sealing parameters, locking parameters, initial eccentricity, and initial angular deviation.

[0038] The knowledge-accumulating agent 208 is used to write successful strategies, failure cases, corresponding parameter ranges, and transfer rules into the knowledge base module 109. Preferably, the knowledge-accumulating agent 208 can also perform feature extraction and indexing on different connector samples to support subsequent strategy transfer.

[0039] 3. Construction of Connector Parametric Digital Objects like Figure 3 As shown, in a preferred embodiment, the parametric modeling module 102 constructs the underwater wet-plug connector 301 as a parametric digital object. The digital object includes at least a mating structure 302, a guiding structure 303, a sealing structure 304, and a locking structure 305.

[0040] The mating parameter set 306 preferably includes the pin diameter, the socket diameter, and the mating stroke. The mating parameter set 307 is used to determine the contact relationship and mating resistance characteristics during the mating process.

[0041] The guide parameter set 307 preferably includes radial alignment deviation, angular alignment deviation, and rotational alignment deviation. The guide parameter set 306 is used to determine the capture capability and alignment tolerance during the approach and alignment phases.

[0042] The sealing parameter set 308 preferably includes sealing resistance, seal thickness, and insertion / extraction speed. The sealing parameter set 308 is used to determine the resistance, deformation, and recoverability during the seal penetration phase.

[0043] The locking parameter set 309 preferably includes locking stroke, mating force, and disconnecting force. The locking parameter set 309 is used to determine the triggering conditions and locking reliability of the locking action.

[0044] In a preferred embodiment, the parameterized digital object may further include one or more of the following features: a two-stage oil-filled pressure balance 310, a penetrable self-sealing 311, and an automatic interlocking docking 312. These features can serve as important inputs to the structural understanding agent 203 and the physical modeling agent 204, enabling more realistic simulation of the connector's autonomous docking behavior in an underwater environment.

[0045] 4. Construction of Simulation Scenario for Autonomous Docking Process like Figure 4As shown, in a preferred embodiment, the scenario construction module 103 constructs the following autonomous docking process simulation scenario.

[0046] In approach phase 401, the actuator end, carrying the plug end, gradually approaches the socket end, but contact has not yet occurred. This phase mainly focuses on the relative positional relationship, actuator path, and approach speed.

[0047] In alignment phase 402, the plug and socket ends are spatially aligned using attitude correction actions. This phase primarily focuses on axial misalignment, lateral misalignment, angular misalignment, and guide segment capture capability.

[0048] In the insertion phase 403, the plug end begins to enter the socket end, initial contact occurs, and insertion force is generated. This phase primarily focuses on the contact point location, contact sequence, magnitude and trend of the insertion force, and whether jamming occurs.

[0049] During the penetration stage 404, the plug end penetrates the sealing area, causing localized deformation of the sealing structure. This stage primarily focuses on the amount of seal deformation, localized seal stress, penetration resistance, and whether overload or instability occurs.

[0050] In the abnormal rollback and retry phase 405, when a jam, excessive insertion force, locking failure, or alignment failure occurs during strategy execution, the system performs a rollback action, realigns, and tries again. This phase is mainly used to evaluate the strategy's resilience and robustness.

[0051] In the locking phase 406, the connector completes the preset insertion depth and triggers the locking action. This phase mainly focuses on the locking trigger position, whether locking is completed, whether the locking stroke meets the requirements, and the consistency and stability of the locking action.

[0052] In each of the above stages, the simulation execution module 105 outputs virtual state quantities 407. Preferably, the virtual state quantities 407 include at least relative pose, contact state, insertion force, seal deformation, and locking state. The virtual state quantities 407 can be further provided to the strategy generation module 107, the parameter optimization module 108, and the criticism and verification module 106.

[0053] 5. Simulation-Criticism-Strategy Generation-Optimization Closed-Loop Execution Process like Figure 5 As shown, the method of the present invention preferably includes the following execution flow. First, the task input module 101 receives task input to form a task definition 501. The task definition 501 may include the target connector model, connector geometry model, environmental pressure, maximum allowable insertion force, target success rate, and actuator speed constraints, etc.

[0054] According to task definition 501, parametric modeling module 102 constructs a connector parametric digital object; then, based on the parametric digital object and environmental conditions, a scenario 503 is constructed, the scenario including at least one or more of ideal scenarios and perturbed scenarios.

[0055] The simulation execution module 105 performs process simulation 504 under the cooperation of multiple agents and outputs a sequence of virtual state variables. The result criticism agent 206 criticizes the simulation results 505. If it identifies a violation of constraints, a risk of engineering unavailability, or insufficient robustness, it outputs criticism to the parameter correction step 512.

[0056] The strategy generation module 107 generates candidate strategies 506 based on simulation results and task objectives. The candidate strategies preferably include a set of phased actions, such as approach paths, attitude correction, insertion speed control, micro-rotation actions, and locking actions. The parameter optimization module 108 optimizes the parameters in the candidate strategies 507. Preferably, parameter optimization can employ reinforcement learning, Bayesian optimization, evolutionary search, hierarchical search, or a combination thereof, but this invention does not limit the specific type of optimization algorithm.

[0057] After obtaining the optimization results, the optimized parameters are re-input into the simulation execution module 105 to perform re-simulation and iteration 508. In one embodiment, the system first determines whether the constraints are satisfied 510. The constraints include at least the maximum insertion force constraint, the sealing deformation constraint, the locking completion constraint, and the docking success rate constraint. If not satisfied, the system proceeds to the parameter correction step 512.

[0058] Once the constraints are satisfied, the next step is to determine whether the strategy is usable (511). Here, "usable" preferably refers to usability in an engineering-implementable sense, rather than simply successful value solving. If unusable, proceed to the parameter correction step (512). The parameter correction step (512) corrects the structural parameters, action parameters, or scenario parameters based on the feedback results, and then returns to the process simulation (504) or candidate strategy generation (506).

[0059] Once the strategy passes the constraint and availability checks, the optimal strategy 509 is output. The output should ideally include the optimal action sequence, parameter combinations, success rate assessment, applicable operating conditions, and recommended usage boundaries.

[0060] In this invention, the typical evaluation index 513 preferably includes success rate, docking time, peak insertion force, locking completion rate and robustness.

[0061] 6. Strategy migration and knowledge base update like Figure 6As shown, the present invention further provides a strategy migration and knowledge base update mechanism. For connector model A601 that has completed simulation and optimization, the system writes its corresponding successful strategy set 602 and failure case set 603 into the knowledge base 604.

[0062] The knowledge-accumulating intelligent agent 208 performs feature extraction and indexing on connector samples 605. The extracted features include, but are not limited to, geometric similarity features, guide structure features, sealing structure features, locking structure features, parameter range features, and success strategy labels.

[0063] When faced with a new connector model B606 or connector model C607, the system performs a similarity assessment 610. If model B or model C achieves a preset similarity with existing samples in terms of geometry, guidance, sealing, or locking features, then the strategy template 608 and initial parameter suggestions 609 can be retrieved from the knowledge base 604.

[0064] Strategy template 608 is used to provide a transferable action sequence framework for new connector models; initial parameter suggestions 609 are used to provide a better initial parameter range for new connector models, thereby reducing the number of random searches and optimization iterations.

[0065] For new strategy results subsequently generated for connector models B606 and C607, the system writes the new successful and failed samples back to the knowledge base 604 through update feedback 611, thereby continuously expanding the coverage of the knowledge base. In a preferred embodiment, the strategy migration and knowledge base update mechanism can achieve one or more of the following effects: rapid migration, reduced iteration count, and improved initial success rate.

[0066] The following example is provided in this embodiment. Example 1: For a certain type of underwater wet-plug connector, its three-dimensional geometric model, material parameters, and task conditions are first imported; the parametric modeling module 102 generates a set of guiding parameters, a set of mating parameters, a set of sealing parameters, and a set of locking parameters; the scene construction module 103 generates an ideal scene, an eccentric scene, and an angular deviation scene; the large model multi-agent collaboration module 104 completes task decomposition, modeling, and simulation scheduling; the simulation execution module 105 outputs the relative pose, contact state, insertion force, and locking state; the result criticism agent 206 identifies high insertion force and locking failure samples; the parameter optimization agent 207 optimizes the approach speed, micro-rotation angle, and locking threshold; finally, the optimal docking strategy that meets the requirements of success rate and peak insertion force is output.

[0067] Example 2: For underwater wet-plug connectors, the penetrable self-sealing function is one of the basic functions for achieving insertion and separation in wet environments. The penetrable self-sealing function can be implemented by different sealing structure forms. Based on this, while optimizing the strategy, the parameter optimization module 108 can also jointly optimize sealing-related parameters such as sealing thickness, sealing angle, and allowable deformation, thereby realizing the joint design of action parameters and sealing structure parameters.

[0068] Example 3: After the knowledge of connector model A has been accumulated, when a new connector model B is input, the system first uses similarity assessment 610 to determine whether it can call the existing strategy template. If the similarity reaches the preset threshold, the system directly extracts the strategy template 608 and the initial parameter suggestion 609 from the knowledge base 604 as the initial solution, and then performs a small number of iterative optimizations to shorten the strategy generation time.

[0069] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for autonomous docking of underwater wet-plug connectors, characterized in that: Includes the following steps: Step 1: Obtain the structural model, material parameters, environmental parameters, execution constraints, and target indicators of the target underwater wet-plug connector; Step 2: Construct a connector parameterized digital object based on the structural model from Step 1. The parameterized digital object includes at least one or more of the following: mating structure parameters, guiding structure parameters, sealing structure parameters, and locking structure parameters. Step 3: Construct a simulation scenario for the autonomous docking process of the connector. The simulation scenario shall include at least one or more of the following scenarios: approach scenario, alignment scenario, insertion scenario, seal penetration scenario, abnormal retreat and retry scenario, and locking scenario. Step 4: Set up a large-scale model multi-agent collaborative module to decompose the autonomous docking task and perform structural understanding, physical modeling, simulation execution, result critique, strategy generation, and parameter optimization respectively. Step 5: Perform an autonomous docking process simulation to obtain the corresponding virtual state variable sequence; Step 6: Generate candidate docking strategies based on the virtual state variable sequence; Step 7: Critically evaluate and verify the candidate docking strategies, and determine whether they meet the preset constraints and target indicators. When the candidate docking strategy does not meet the preset constraints and / or target indicators, optimize one or more of the action parameters, structural parameters, and operating condition parameters, and perform resimulation and iteration. When the candidate docking strategy meets the preset constraints and / or target indicators, output the optimal docking strategy, optimized parameters, and applicable operating conditions that meet the target requirements. Based on the output optimal docking strategy, optimized parameters, and applicable operating conditions, complete the autonomous docking of the underwater wet-plug connector.

2. An autonomous docking system for underwater wet-plug connectors, equipped with the autonomous docking method for underwater wet-plug connectors as described in claim 1, characterized in that: include: The task input module is used to execute step one to receive input information related to the autonomous docking task; The parametric modeling module is used to perform step two, converting the input connector model into a parametric digital object that can be used for simulation and optimization. The scenario construction module is used to execute step three to construct different simulation scenarios of the docking process based on the autonomous docking task definition. The large-scale model multi-agent collaboration module is used to execute step four to organize multiple functional agents to decompose, reason, model, critique, and optimize the task; The simulation execution module is used to execute step five, which simulates the autonomous docking process based on the parameterized digital objects and the scene construction results. The critique and verification module is used to perform step six to judge the engineering and physical rationality of the simulation results and candidate strategies; The strategy generation module is used to execute step seven to generate candidate docking strategies and the optimal docking strategy based on the task objective, scenario information and simulation results. The parameter optimization module is used to execute step seven to optimize one or more of the motion parameters, structural parameters, and operating parameters according to a preset objective function. The knowledge base module is used to perform step seven to store successful strategies, failure cases, parameter ranges, applicable scenarios, and migration rules. The output results module is used to execute step seven to output the optimal docking strategy, optimization parameters, and applicable operating conditions.

3. The underwater wet-plug connector autonomous docking system according to claim 2, characterized in that: The large-scale model multi-agent collaborative module includes: The Collaboration Manager agent is used to dynamically allocate task order and invocation relationships based on the current task context. A task planning agent is used to break down the input autonomous docking task into several sub-tasks; A structural understanding agent is used to perform semantic parsing of the geometry and assembly relationships of connectors; A physical modeling agent is used to construct a physical model that adapts to the autonomous docking process based on the results of structural analysis. The simulation execution agent is used to call the simulation execution module to solve the simulation task based on the simulation task description generated by the physical modeling agent, and then return the simulation results. The result critique agent is used to critique and verify the simulation results; The parameter optimization agent is used to optimize relevant parameters based on the criticism results and target indicators given by the result criticism agent 206. The knowledge-accumulating intelligent agent is used to write successful strategies, failure cases, corresponding parameter ranges, and migration rules into the knowledge base module.

4. The underwater wet-plug connector autonomous docking system according to claim 3, characterized in that: The parametric modeling module transforms the input connector model into a parametric digital object that can be used for simulation and optimization. Specifically: The mating parameter set, including pin diameter, socket diameter, and mating stroke, is used to determine the contact relationship and mating resistance characteristics during the mating process; The set of guidance parameters, including radial alignment deviation, angular alignment deviation, and rotational alignment deviation, is used to determine the capture capability and alignment tolerance during the approach and alignment phases; A set of sealing parameters, including sealing resistance, seal thickness, and insertion / extraction speed, is used to determine the resistance, deformation, and recoverability during the seal penetration phase. The locking parameter set, including locking stroke, docking force, and disconnecting force, is used to determine the triggering conditions and locking reliability of the locking action.

5. The underwater wet-plug connector autonomous docking system according to claim 4, characterized in that: The scenario construction module builds a simulation scenario for the autonomous docking process, including the following stages: During the approach phase, the end of the actuator carrying the plug gradually approaches the socket end, but contact has not yet occurred; During the alignment phase, the plug and socket ends are spatially aligned using attitude correction actions. During the insertion phase, the plug end begins to enter the socket end, initial contact occurs, and insertion force is generated. During the sealing penetration stage, the plug end penetrates the sealing area, causing local deformation of the sealing structure. During the abnormal rollback and retry phase, if the strategy execution process encounters issues such as jamming, excessive insertion force, locking failure, or alignment failure, the system will perform a rollback action, realign, and try again. During the locking phase, the connector completes the preset insertion depth and triggers the locking action.

Images