Six-degree-of-freedom cooperative control ship-based unmanned aerial vehicle dynamic deck recovery system

Abstract

The invention discloses a six-degree-of-freedom cooperative control ship-based unmanned aerial vehicle dynamic deck recovery system, and belongs to the technical field of ship-based unmanned aerial vehicle recovery. The system comprises an information acquisition and agent networking module for acquiring various types of data, constructing a network and unifying a format; the motion prediction and confidence evaluation module constructs a model prediction trend, calculates confidence and transmits data in a linkage manner; the consensus decision and link reconstruction module realizes information interaction and reconstructs a framework; the dynamic compensation and attitude calibration module is used for adjusting the equipment calibration attitude; the flexible recovery and circulation preparation module completes recovery and circulation preparation; according to the method, the decentralized navigation agent network is constructed, so that the problems of data isomerism and the like are solved, and the data processing accuracy and timeliness are improved; the attitude deviation is accurately corrected through a machine-ship linkage adjustment instruction generation algorithm and the like; the control precision is improved through full-process collaborative design, efficient connection such as recycling is achieved, the system operation flexibility and continuity are enhanced, a closed loop is formed, and the overall efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of shipborne unmanned aerial vehicle (UAV) recovery technology, specifically a shipborne UAV dynamic deck recovery system with six degrees of freedom collaborative control. Background Technology

[0002] The integration of multi-agent cooperative control technology with visual perception and multi-source navigation data fusion technology has become an important development direction in the dual-use maritime operations field. Cooperative operations between ships and drones can significantly expand the coverage and efficiency of maritime missions, playing a crucial role in tasks such as maritime reconnaissance, environmental monitoring, and communication relay. The maritime environment is characterized by complex and ever-changing disturbances; ships exhibit multi-degree-of-freedom motion, and the flight attitude of drones is easily affected indirectly by sea winds and waves. This places stringent requirements on motion matching, navigation accuracy, and recovery reliability between ships and drones. To ensure the stability of cooperative operations, it is necessary to integrate various sensing devices such as millimeter-wave arrays, visual units, and navigation modules to acquire motion data and environmental information from ships and drones. Data processing, decision generation, and execution control are achieved through collaborative interaction between agents. Currently, related technologies are evolving towards multi-source data integration and intelligent autonomous decision-making. However, issues such as disturbance adaptation, data synchronization, and cooperative response speed caused by complex maritime conditions still require more sophisticated technical solutions to meet the demands of high-precision and high-reliability cooperative operations.

[0003] Traditional ship-UAV collaborative control systems often employ a centralized network architecture, with decision-making and data transmission overly reliant on core nodes. Failure of a core node or disruption of the link can severely impact the stability of the entire system, potentially even leading to the interruption of collaborative operations. In the data processing stage, the raw data from different sensing devices lacks unified calibration standards and format specifications, making data consistency difficult to guarantee. Furthermore, the fusion of multi-source data is relatively simple, failing to fully consider the dynamic changes in the device's own state and environmental disturbances, resulting in insufficient accuracy in navigation decisions. The confidence assessment mechanism is not comprehensive enough, relying on a single dimension to judge data validity, making it difficult to effectively eliminate low-confidence data, thus affecting the scientific nature of collaborative decisions. The dynamic compensation and attitude calibration stages lack deep integration with navigation data and environmental disturbances, resulting in delayed adjustment responses and an inability to quickly adapt to the real-time movement changes of ships and UAVs. In addition, the UAV recovery process lacks an effective flexible buffer mechanism, the fixing method is unreliable, and the payload replacement process is cumbersome and requires system interruption, severely impacting operational continuity and efficiency. Overall, it is difficult to adapt to the collaborative operation requirements in complex maritime environments. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a six-degree-of-freedom collaborative control system for shipborne unmanned aerial vehicles (UAVs) with dynamic deck recovery. It integrates multi-source sensing data to construct a decentralized network, achieves motion prediction and credibility assessment through a ship-UAV collaborative dynamics model and a multi-dimensional confidence fusion algorithm, completes decision-making and dynamic link reconstruction based on a weighted voting consensus algorithm, and achieves precise dynamic compensation and attitude calibration through a ship-UAV linkage adjustment mechanism. Combined with flexible recovery, locking and fixing, and fast load changing functions, it ensures efficient operation cycles. The system can dynamically adapt to complex disturbances at sea, solving problems such as data heterogeneity, response lag, and insufficient operational continuity in traditional collaborative control. It achieves high-precision and high-reliability collaboration in maritime missions, significantly improving operational efficiency and environmental adaptability.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a six-degree-of-freedom cooperative control shipborne unmanned aerial vehicle dynamic deck recovery system, the system comprising: Information Acquisition and Intelligent Agent Networking Module: Collects six-degree-of-freedom motion data of the ship and raw perception data from multiple heterogeneous navigation sensors, completes parameter calibration and status self-check of each navigation intelligent agent, builds a decentralized navigation intelligent agent network, and unifies data format; Motion prediction and confidence assessment module: Based on the collected data, a ship-machine cooperative dynamics model is constructed to predict the ship's motion trend. Based on the output of the model and the data of each intelligent agent, the fusion confidence of each navigation intelligent agent is calculated through a multi-dimensional dynamic confidence fusion algorithm. The motion prediction results, intelligent agent pose data and fusion confidence data are transmitted in a coordinated manner. Consensus Decision and Link Reconstruction Module: Receives motion prediction results, agent pose data and fusion confidence data, realizes autonomous information interaction among agents through a confidence-based weighted voting consensus algorithm, identifies and removes failed nodes, dynamically reconstructs the navigation fusion architecture, generates fusion navigation instructions and calculates the average fusion confidence of all effective agents; Dynamic compensation and attitude calibration module: Combining fused navigation commands, motion prediction results and average fused confidence, the module uses a machine-ship linkage adjustment command generation algorithm to adjust the extension and retraction state of the actuating column of the six-degree-of-freedom dynamic compensation platform, drive the intelligent robotic arm to adjust the attitude of each joint, and calibrate the position of the slide rail. Flexible recovery and recycling preparation module: Receives UAV docking speed data and ship pitch angle data, adjusts the state of the magnetorheological damping device to realize UAV recovery, fixes the UAV through the electronic locking mechanism, initiates the power replenishment process and payload quick change operation, and updates the initial state of the intelligent agent network and data acquisition benchmark.

[0006] Furthermore, the parameter calibration of the information acquisition and intelligent agent networking module includes millimeter-wave array ranging zero-point calibration, visual unit fractal marker coordinate calibration and camera intrinsic parameter calibration, and dual IMU zero-bias calibration. The calibration process is completed by comparing data with a standard attitude reference platform. After calibration, a calibration record is generated and stored. The decentralized navigation intelligent agent network is constructed through an Ethernet switch. When each intelligent agent node registers, it binds its device identifier, communication address, and data type. During the network self-test, the hardware connection status, data transmission link connectivity, and power supply stability of each node are detected.

[0007] Furthermore, in the motion prediction and confidence assessment module, the ship-machine cooperative dynamics model is constructed by integrating the dynamic characteristics of the ship and the UAV, including a ship dynamics sub-model, a UAV dynamics sub-model, and a ship-machine coupling mechanism. The ship dynamics sub-model is based on the ship's motion equations, incorporates wave disturbance and ship structural characteristics, inputs the ship's six-degree-of-freedom original motion data, and outputs the short-term motion attitude change trend of the ship. The UAV dynamics sub-model combines the UAV's mass, moment of inertia, and aerodynamic characteristic parameters, inputs the UAV's real-time flight status data, and outputs the UAV's inertial response characteristic data. The ship-machine coupling mechanism, through a data synchronization interface, fuses the outputs of the ship dynamics sub-model and the UAV dynamics sub-model in real time to construct a unified dynamic mapping relationship. The model output data is used for calculating the fusion confidence and generating motion prediction results.

[0008] Furthermore, in the motion prediction and confidence assessment module, the mathematical expression of the multi-dimensional dynamic confidence fusion algorithm is: ,in For the first A navigation agent at any time The dynamic fusion confidence level , , For dynamic adaptive weights, For the first The basic confidence level of a navigation agent itself. For the first The basic confidence level of a navigation agent itself. For the first The number of neighboring effective agents of each navigation agent. For the first The set of neighboring agents of a navigation agent, For the first A navigation agent at any time The pose calculation results, For the first A navigation agent at any time The pose calculation results, It is the minimum value. This is the ship's hull disturbance adaptation coefficient.

[0009] Furthermore, in the consensus decision-making and link reconstruction module, after data reception, timestamp alignment is first performed. Using the system's reference timeline as a unified reference, linear interpolation is used to correct the temporal deviation of data from different navigation agents. A fusion confidence threshold of 0.3 is set. For navigation agents with a fusion confidence lower than the threshold, they are marked as suspected failure nodes and a secondary confirmation process is triggered: three adjacent navigation agents designated by the system verify the continuity of the suspected node's data and the consistency of its pose data with the verification node's own data within the most recent preset time period. When all three verification nodes report verification failure, the node is determined to be a failure node and is removed. During navigation fusion architecture reconstruction, when only a single navigation agent fails, the remaining navigation agents are reassigned weights according to their respective fusion confidence proportions, with the total weight allocation remaining at 1. When multiple navigation agents are disturbed, the top three navigation agents with the highest fusion confidence are extracted first to construct the core navigation link. The total fusion confidence of the three agents in the core navigation link is not less than 2.0.

[0010] Furthermore, the execution process of the credibility-based weighted voting consensus algorithm is as follows: During the information interaction phase, each navigation agent actively sends its own pose data, fusion confidence data, and data acquisition timestamps to all navigation agents in its neighborhood; after receiving the neighborhood data, each navigation agent performs normalization processing based on the fusion confidence of the other party to obtain the voting weight of the corresponding neighborhood navigation agent, and the sum of the voting weights of all neighborhood navigation agents is 1; during the voting phase, each navigation agent votes on the target pose parameters based on its own solution results and neighborhood data, and the voting result is multiplied by the corresponding weight to obtain the weighted voting value; during the statistical phase, invalid votes with weighted voting values ​​lower than 0.05 are removed, and the remaining valid weighted voting values ​​are summed to obtain the pose parameter consensus result; finally, combining the pose consensus results of all valid navigation agents, a unified high-precision fused navigation command is generated through mean fusion, and the average fusion confidence is calculated according to the proportion of fusion confidence of each valid navigation agent.

[0011] Furthermore, in the dynamic compensation and attitude calibration module, the mathematical expression for the machine-ship linkage adjustment command generation algorithm is: ,in For a moment The set of linkage adjustment commands, The confidence level weighting coefficient is... For all effective navigation agents at time The average fusion confidence, To adjust the gain of the six-degree-of-freedom dynamic compensation platform, For a moment Attitude deviation of the takeoff and landing reference plane, For the attitude adjustment gain of intelligent robotic arms, For a moment The alignment deviation between the centerline of the slide rail and the flight vector of the UAV. To adjust for sea state conditions, For a moment The deviation in the docking speed of the drone This is the time variation of the velocity deviation. For a moment The real-time value of the ship's pitch angle.

[0012] Furthermore, in the dynamic compensation and attitude calibration module, the six-degree-of-freedom dynamic compensation platform adopts an improved parallel configuration, with the upper and lower end plates being symmetrical rigid structures. The actuating columns are connected to the upper and lower end plates through spherical joints, and the actuating columns are arranged symmetrically in pairs to form a spatial triangular configuration. The extension and retraction of the actuating columns are controlled by a servo drive unit. The intelligent robotic arm adopts a serial configuration, including a rotary joint, a front-end telescopic mechanism, and an auxiliary fine-tuning joint. The attitude of the slide rail is adjusted through the coordinated action of multiple joints. During the slide rail calibration process, feedback data from the visual sensor is received in real time. When a deviation is detected between the center line of the slide rail and the flight vector of the UAV, the fine-tuning parameters are immediately recalculated based on the linkage adjustment command, and a secondary attitude correction is initiated to ensure that the slide rail and the flight path of the UAV remain matched.

[0013] Furthermore, in the flexible recycling and recycling preparation module, the magnetorheological damping device consists of a magnetorheological fluid cavity and a carbon fiber beam segment, arranged in a specific section at the end of the slide rail. The viscosity of the magnetorheological fluid is changed by adjusting the magnetic field strength. The electronically controlled locking mechanism includes an electromagnetic lock and a mechanical positioning pin. After the UAV comes to rest, it is first initially fixed by the electromagnetic lock, and then limited a second time by the mechanical positioning pin. After locking, the locking status is detected by the sensor.

[0014] Furthermore, in the flexible recycling and recycling preparation module, the load quick-change interface adopts a modular structure, including a power bus, a data bus, and a mechanical positioning structure. When changing the load, the load tray is pulled out to replace the functional module, and after being pushed back, it is automatically guided to the correct position by the mechanical positioning structure, and the power and data interfaces are connected simultaneously. The entire replacement process does not require a system power outage. Beneficial effects

[0015] Compared with existing technologies, this six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system has the following advantages: I. This invention constructs a decentralized navigation agent network, integrates multi-source sensing data, and completes parameter calibration and state self-checking. While unifying the data format, it ensures network connectivity and stability. Relying on the ship-machine cooperative dynamics model, it accurately predicts motion trends. Combined with a multi-dimensional dynamic confidence fusion algorithm, it comprehensively evaluates the credibility of each agent. Through a weighted voting consensus algorithm, it realizes autonomous information interaction and consensus decision-making among agents, dynamically eliminates low-confidence nodes, and reconstructs the navigation architecture. This design effectively solves the problems of data heterogeneity and low collaborative efficiency in traditional systems, improves the accuracy and timeliness of data processing, and enables the system to maintain efficient collaboration even in complex interference environments. It provides solid data support for the generation of subsequent control commands, ensures the stability and reliability of overall operation, adapts to various dynamically changing operating scenarios, and significantly reduces the impact of environmental disturbances on the collaborative control effect.

[0016] Second, this invention utilizes a machine-ship linkage adjustment command generation algorithm, linking a six-degree-of-freedom dynamic compensation platform, an intelligent robotic arm, and a slide rail calibration mechanism. It responds in real-time to motion prediction results and average fusion confidence levels, accurately correcting attitude deviations and positional misalignments. The flexible recovery system achieves stable drone recovery using a magnetorheological damping device, and achieves double fixation through an electronically controlled locking mechanism. Combined with a modular payload quick-change interface and power replenishment process, it rapidly updates the initial state of the intelligent agent network and data acquisition benchmarks. This end-to-end collaborative design not only improves the control accuracy of the actuators and reduces the impact of dynamic interference on operations, but also achieves efficient integration of recovery, power replenishment, and payload replacement. Cyclic preparation can be completed without system interruption, significantly enhancing the system's operational flexibility and continuous operation capability. It meets the high-precision operation requirements under complex conditions, enabling multi-agent collaboration to form a closed loop from data processing to execution, comprehensively improving overall operational efficiency and reliability.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 A flowchart of a six-degree-of-freedom collaborative control shipborne unmanned aerial vehicle (UAV) dynamic deck recovery system; Figure 2This diagram illustrates the data transmission between modules of a shipborne UAV dynamic deck recovery system with six degrees of freedom collaborative control. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below. Example

[0021] Drone recovery and mission switching in maritime law enforcement patrol scenarios under sea state 6 Coast Guard vessels conducted long-range patrol and evidence collection missions in sea state 6. They used an air-sea rotor system to autonomously recover drones, quickly change payloads, and then release them. This enabled them to conduct an average of 50 high-frequency law enforcement and evidence collection missions per day, with recovery accuracy controlled within ±5 centimeters, an autonomous recovery success rate of over 90%, and a payload switching time of no more than 30 seconds. The entire process required no human intervention.

[0022] Information Acquisition and Intelligent Agent Networking: After system startup, the information acquisition and intelligent agent networking module synchronously collects six-degree-of-freedom motion data from the law enforcement vessel, as well as raw perception data from heterogeneous navigation sensors such as millimeter-wave arrays, vision units, and dual IMUs. This comprehensively captures the ship's pitch, roll, and sway motion states, along with real-time perception information from each sensor. Subsequently, it completes millimeter-wave array ranging zero-point calibration, vision unit fractal mark coordinate calibration and camera intrinsic parameter calibration, and dual IMU zero-bias calibration. Calibration accuracy is ensured by comparing data with a standard attitude reference platform, and calibration records are stored in real-time for subsequent traceability. Simultaneously, a decentralized navigation intelligent agent network is constructed. Each intelligent agent node binds its device identifier, communication address, and data type during registration, achieving efficient connection through an Ethernet switch. During network self-testing, the hardware connection status of each node, data transmission link connectivity, and power supply stability are comprehensively checked to eliminate potential fault points. After standardizing the data format, efficient transmission of multi-source data to subsequent processing modules is ensured, as shown in the figure. Figure 1 As shown.

[0023] Motion Prediction and Confidence Assessment: The motion prediction and confidence assessment module receives multi-source data, integrates the dynamic characteristics of the law enforcement vessel and the UAV, and constructs a ship-aircraft cooperative dynamics model that includes a ship dynamics sub-model, a UAV dynamics sub-model, and a ship-aircraft coupling mechanism. The ship dynamics sub-model incorporates the effects of wave disturbances under sea state 6 and the ship's structural characteristics. By inputting the ship's six-degree-of-freedom raw motion data, it accurately outputs the short-term trend of the ship's motion attitude. The UAV dynamics sub-model combines the mass, moment of inertia, and aerodynamic parameters of the UAV performing the evidence collection task, inputting real-time flight status data of the UAV, and accurately outputs the UAV's inertial response characteristic data. The ship-aircraft coupling mechanism fuses the outputs of the two sub-models in real time through a data synchronization interface, forming a unified dynamic mapping relationship. Based on the model output and the data of each agent, the fusion confidence of each navigation agent is calculated using a multi-dimensional dynamic confidence fusion algorithm. The mathematical expression of the multi-dimensional dynamic confidence fusion algorithm is: ,in For the first A navigation agent at any time The dynamic fusion confidence level , , For dynamic adaptive weights, For the first The basic confidence level of a navigation agent itself. For the first The number of neighboring effective agents of each navigation agent. For the first The set of neighboring agents of a navigation agent, For the first A navigation agent at any time The pose calculation results, For the first A navigation agent at any time The pose calculation results, It is the minimum value. To adapt to ship disturbances, the reliability of data from each agent is comprehensively evaluated. Simultaneously, motion prediction results, agent pose data, and fused confidence data are transmitted to the consensus decision-making module to provide data support for subsequent decisions.

[0024] Consensus Decision-Making and Link Reconstruction: After receiving data, the consensus decision-making and link reconstruction module uses the system's baseline timeline as a unified reference and employs linear interpolation to correct temporal deviations in the data of different navigation agents, ensuring that all data remain consistent across time. The system sets a fusion confidence threshold of 0.3. Navigation agents with a fusion confidence score below this threshold are marked as suspected failure nodes, triggering a secondary confirmation process. Three adjacent navigation agents designated by the system verify the continuity of the suspected node's data within the most recent preset time period and the consistency of its pose data with the verification node's own data. If all three verification nodes report verification failure, the node is determined to be a failure node and removed to prevent erroneous data from affecting system decisions. Since some agents may be disturbed under sea state 6, the system prioritizes extracting the top three navigation agents with the highest fusion confidence scores to construct the core navigation link, ensuring that the sum of the fusion confidence scores of the three agents is not less than 2.0, guaranteeing the stability of the core navigation functions. Subsequently, an autonomous information exchange among the agents is achieved through a credibility-based weighted voting consensus algorithm. Each agent sends its own pose data, fusion confidence data, and data acquisition timestamps to all agents in its neighborhood. After receiving neighborhood data, it performs normalization processing based on the fusion confidence of the other party to obtain voting weights. Based on its own calculation results and neighborhood data, it votes on the target pose parameters. After removing invalid votes with weighted voting values ​​lower than 0.05, the remaining valid weighted voting values ​​are summed to obtain the pose parameter consensus result. Then, a unified high-precision fusion navigation command is generated through mean fusion. At the same time, the average fusion confidence is calculated according to the proportion of fusion confidence of each valid navigation agent, providing a basis for attitude calibration.

[0025] Dynamic Compensation and Attitude Calibration: The dynamic compensation and attitude calibration module combines the generated fused navigation commands, motion prediction results, and average fused confidence scores to accurately calculate adjustment parameters using a machine-ship coordinated adjustment command generation algorithm. The mathematical expression of the machine-ship coordinated adjustment command generation algorithm is as follows: ,in For a moment The set of linkage adjustment commands, The confidence level weighting coefficient is... For all effective navigation agents at time The average fusion confidence, To adjust the gain of the six-degree-of-freedom dynamic compensation platform, For a moment Attitude deviation of the takeoff and landing reference plane, For the attitude adjustment gain of intelligent robotic arms, For a moment The alignment deviation between the centerline of the slide rail and the flight vector of the UAV. To adjust for sea state conditions, For a moment The deviation in the docking speed of the drone This is the time variation of the velocity deviation. For a moment The system calculates the real-time pitch angle of the ship. Based on the calculation results, it adjusts the extension and retraction of the actuators on the six-degree-of-freedom dynamic compensation platform. This platform adopts an improved parallel configuration with symmetrical rigid structures on the upper and lower end plates. The actuators are arranged in pairs in a symmetrical spatial triangular configuration. The extension and retraction of the actuators are precisely controlled by a servo drive unit to compensate for pitch, roll, and other motion disturbances of the ship in sea state 6, providing a stable reference plane for UAV recovery. At the same time, it drives the intelligent robotic arm to adjust the attitude of each joint. This robotic arm adopts a serial configuration and includes a rotary joint, a front extension mechanism, and auxiliary fine-tuning joints. Through the coordinated action of multiple joints, it calibrates the position of the slide rail to ensure that the center line of the slide rail is precisely aligned with the flight vector of the UAV, improving the recovery alignment accuracy.

[0026] Flexible Recovery and Recycling Preparation: After the UAV arrives over the law enforcement vessel, the flexible recovery and recycling preparation module receives real-time data on the UAV's docking speed and the vessel's pitch angle. It adjusts the magnetic field strength of the magnetorheological damping device, composed of a magnetorheological fluid cavity and carbon fiber beams. By altering the viscosity of the magnetorheological fluid, it achieves flexible buffering, reducing the peak impact during UAV recovery to less than 60% of that of a conventional fixed arm, preventing damage to the UAV and completing autonomous landing and recovery. After recovery, the electronic locking mechanism first initially secures the UAV with an electromagnetic lock, then uses mechanical positioning pins for secondary positioning. After locking, sensors detect the locking status to ensure the UAV is securely fixed. The system then automatically initiates the power replenishment process. Operators use simple tools to quickly switch between evidence collection and reconnaissance payloads within 30 seconds using a sliding-rail modular payload bay, without requiring a power outage, ensuring rapid mission switching. Finally, the system updates the initial state of the intelligent agent network and data acquisition baselines to prepare for the next UAV release, supporting the continuous operation of 50 high-frequency law enforcement evidence collection missions per day.

[0027] In summary, in a sea state 6 maritime law enforcement patrol scenario, the air-sea rotor system efficiently completes autonomous drone recovery and mission switching through the coordinated operation of five modules. The information acquisition and intelligent agent networking module solidifies the data foundation; the motion prediction and confidence assessment module accurately outputs situational data; the consensus decision-making and link reconstruction module ensures command reliability; the dynamic compensation and attitude calibration module achieves centimeter-level recovery accuracy; and the flexible recovery and cyclic preparation module supports 30-second payload quick change. The entire process requires no manual intervention, meeting the core requirements of ±5 cm recovery accuracy and over 90% autonomous recovery success rate, while also achieving a high-frequency operation of 50 sorties per day. This fully adapts to the stringent demands of far-sea law enforcement and evidence collection, demonstrating strong adaptability to high sea states and mission continuity. Example

[0028] Rapid deployment and inspection of drones in deep-sea wind power operation and maintenance scenarios.

[0029] Deep-sea wind farms require frequent inspections of the turbine blades. Maintenance vessels equipped with air-sea rotor systems enable drones to launch in seconds under sea states 4-6, complete inspections within 2 minutes, recover the drones, and reload within 30 seconds, minimizing downtime losses and the risks of manual climbing. Navigation continuity during inspections exceeds 99.8%. Figure 2 As shown.

[0030] Information Acquisition and Intelligent Agent Networking: After the maintenance vessel docks near the target wind turbine, the system activates the information acquisition and intelligent agent networking module. This module simultaneously collects six-degree-of-freedom motion data from the maintenance vessel, as well as raw sensor data from heterogeneous navigation sensors such as millimeter-wave arrays, vision units, and dual IMUs, comprehensively understanding the ship's motion status and sensor perception information. Subsequently, it completes millimeter-wave array ranging zero-point calibration, vision unit fractal marker coordinate calibration and camera intrinsic parameter calibration, and dual IMU zero-bias calibration. Calibration accuracy is ensured by comparing data with standard attitude reference station data, and calibration records are stored for future reference. Simultaneously, a decentralized navigation intelligent agent network is constructed. Each intelligent agent node binds its device identifier, communication address, and inspection-related data types during registration. Efficient node connection is achieved using Ethernet switches. During network self-checking, the hardware connection status of each node, data transmission link connectivity, and power supply stability are comprehensively verified, eliminating link faults and power supply hazards. After standardizing the data format, it ensures smooth transmission of multi-source data to subsequent modules, providing a data foundation for subsequent operations.

[0031] Motion Prediction and Confidence Assessment: After receiving the collected data, the motion prediction and confidence assessment module integrates the dynamic characteristics of the maintenance vessel and the inspection drone to construct a ship-machine collaborative dynamic model, which includes a ship dynamics sub-model, a drone dynamics sub-model, and a ship-machine coupling mechanism. The ship dynamics sub-model incorporates wave disturbances around the wind turbine and the ship's structural characteristics, inputting the ship's original six-degree-of-freedom motion data and accurately outputting the short-term trend of the ship's motion attitude. The drone dynamics sub-model combines the inspection drone's mass (within the range of 8-18 kg), moment of inertia, and aerodynamic parameters, inputting the drone's real-time state data and accurately outputting inertial response characteristic data. The ship-machine coupling mechanism fuses the outputs of the two sub-models in real time through a data synchronization interface, forming a unified dynamic mapping relationship. Based on the model output and the data of each agent, the fusion confidence of each navigation agent is calculated using a multi-dimensional dynamic confidence fusion algorithm. The mathematical expression of the multi-dimensional dynamic confidence fusion algorithm is: ,in For the first A navigation agent at any time The dynamic fusion confidence level , , For dynamic adaptive weights, For the first The basic confidence level of a navigation agent itself. For the first The number of neighboring effective agents of each navigation agent. For the first The set of neighboring agents of a navigation agent, For the first A navigation agent at any time The pose calculation results, For the first A navigation agent at any time The pose calculation results, It is the minimum value. To adapt to ship disturbances, the reliability of data from each agent is comprehensively evaluated. Simultaneously, motion prediction results, agent pose data, and fused confidence data are transmitted to the consensus decision-making module, providing accurate data support for catapult and inspection guidance.

[0032] Consensus Decision-Making and Link Reconstruction: After receiving data, the consensus decision-making and link reconstruction module uses linear interpolation to correct the temporal deviations of data from different agents, referencing the system's baseline timeline, ensuring data synchronization across time. The system filters agents based on a fusion confidence threshold of 0.3. For suspected failed nodes below this threshold, a secondary confirmation process is initiated, where three adjacent agents verify data continuity and consistency. If verification fails, the node is removed to prevent erroneous data from interfering with subsequent operations. Since sea conditions are relatively stable in maintenance scenarios, if only a single agent fails, the remaining agents are weighted according to their fusion confidence percentage, with the total weight remaining at 1. If multiple agents are affected, the top three agents are extracted to construct the core navigation link, ensuring their total fusion confidence is not less than 2.0, guaranteeing stable operation of the navigation system. Using a credibility-based weighted voting consensus algorithm, the pose, confidence, and timestamp data of each agent are interacted and normalized to obtain voting weights. After voting, invalid votes are removed and the results are summed to obtain the pose consensus result. The mean is fused to generate high-precision fused navigation commands. At the same time, the average fused confidence of all effective agents is calculated to provide a reliable basis for attitude calibration.

[0033] Dynamic Compensation and Attitude Calibration: The dynamic compensation and attitude calibration module combines fused navigation commands, motion prediction results, and average fused confidence scores to accurately calculate adjustment parameters using a machine-ship coordinated adjustment command generation algorithm. The mathematical expression for this algorithm is: ,in For a moment The set of linkage adjustment commands, The confidence level weighting coefficient is... For all effective navigation agents at time The average fusion confidence, To adjust the gain of the six-degree-of-freedom dynamic compensation platform, For a moment Attitude deviation of the takeoff and landing reference plane, For the attitude adjustment gain of intelligent robotic arms, For a moment The alignment deviation between the centerline of the slide rail and the flight vector of the UAV. To adjust for sea state conditions, For a moment The deviation in the docking speed of the drone This is the time variation of the velocity deviation. For a moment The system monitors the ship's pitch angle in real time. Based on these parameters, the actuation column of the six-degree-of-freedom dynamic compensation platform is adjusted for extension and retraction. The actuation column connects to the upper and lower end plates via spherical joints, forming a spatial triangular configuration with symmetrical cross-sections. This configuration compensates for the heave, roll, and other movements of the maintenance vessel in real time, counteracting disturbances caused by sea conditions and providing a stable reference plane for UAV catapult launches. Simultaneously, it drives the intelligent robotic arm to adjust the attitude of the rotary joint, the front telescopic mechanism, and the auxiliary fine-tuning joints, precisely calibrating the slide rail position to ensure accurate alignment between the slide rail centerline and the UAV's preset flight vector. This guarantees accurate catapult launch direction and lays the foundation for subsequent inspection operations.

[0034] Variable Damping Catapult and Cyclic Operation: The variable damping catapult-buffer integrated device in the flexible recovery and cyclic preparation module switches to catapult mode. The magnetorheological damper is in a low-viscosity state, and the catapult arm releases 12-15g propulsion acceleration within 0.3 seconds, quickly launching the UAV carrying the inspection payload and ensuring that the UAV quickly leaves the ship and enters the inspection route. After the UAV takes off, the optoelectronic-millimeter-wave-inertial multi-mode redundant navigation system is activated. Within a distance of 200 meters, the millimeter-wave array completes the initial ranging guidance. After entering the near field of 50 meters, the vision system takes over, achieving sub-meter level visual navigation and attitude calculation through fractal coding marking. The dual redundant IMU provides high-frequency attitude and acceleration compensation at a frequency of 400Hz. The navigation manager evaluates the link signal-to-noise ratio and obstruction rate in real time. If interference such as sea fog or strong light occurs, the optimal link combination is switched within 50 milliseconds to ensure navigation continuity of over 99.8%, accurately guiding the UAV to complete the wind turbine blade top inspection. After the inspection is completed, the UAV returns to the maintenance vessel. The module receives data on the UAV's docking speed and the ship's pitch angle, adjusts the magnetic field strength of the magnetorheological damping device, and increases the damping to reduce the arm stiffness by 40%, providing secondary buffer energy absorption within the last 10 centimeters of travel, thus completing flexible recovery. After the electronically controlled locking mechanism secures the UAV, the power replenishment process is initiated. The operator pulls out the load tray to replace the inspection load through the drawer-type load quick-change interface, pushes it back, and automatically guides it into position through the mechanical positioning structure, simultaneously completing the power and data interface docking. The entire process is completed within 30 seconds without requiring a system power outage. The system updates the initial state of the intelligent agent network and the data acquisition benchmark, and then the ejection process can be restarted to perform the inspection of the next wind turbine or the delivery of light components, achieving high-frequency cyclical operations and meeting the maintenance needs of deep-sea wind power.

[0035] In summary, in the operation and maintenance (O&M) scenarios of deep-sea wind turbines, the system relies on modular collaboration to accurately match high-frequency inspection needs. The information acquisition and intelligent agent networking module ensures stable data transmission, the motion prediction and consensus decision-making module provides reliable navigation support, the dynamic compensation and attitude calibration module ensures launch accuracy, and the variable damping launch and cyclic operation module achieves second-level takeoff and 30-second rapid load change. The system is adaptable to sea states 4-6, with navigation continuity exceeding 99.8%, and can complete a closed loop of launch-inspection-recovery-load change within 2 minutes, effectively reducing downtime losses and human error. It efficiently meets the O&M needs of deep-sea wind turbines, such as high-frequency inspections and lightweight component delivery, highlighting its technological advantages of lightweight design, high turnaround time, and strong adaptability.

[0036] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A shipborne unmanned aerial vehicle (UAV) dynamic deck recovery system with six degrees of freedom cooperative control, characterized in that, The system includes: Information Acquisition and Intelligent Agent Networking Module: Collects six-degree-of-freedom motion data of the ship and raw perception data from multiple heterogeneous navigation sensors, completes parameter calibration and status self-check of each navigation intelligent agent, builds a decentralized navigation intelligent agent network, and unifies data format; Motion prediction and confidence assessment module: Based on the collected data, a ship-machine cooperative dynamics model is constructed to predict the ship's motion trend. Based on the output of the model and the data of each intelligent agent, the fusion confidence of each navigation intelligent agent is calculated through a multi-dimensional dynamic confidence fusion algorithm. The motion prediction results, intelligent agent pose data and fusion confidence data are transmitted in a coordinated manner. Consensus Decision and Link Reconstruction Module: Receives motion prediction results, agent pose data and fusion confidence data, realizes autonomous information interaction among agents through a confidence-based weighted voting consensus algorithm, identifies and removes failed nodes, dynamically reconstructs the navigation fusion architecture, generates fusion navigation instructions and calculates the average fusion confidence of all effective agents; Dynamic compensation and attitude calibration module: Combining fused navigation commands, motion prediction results and average fused confidence, the module uses a machine-ship linkage adjustment command generation algorithm to adjust the extension and retraction state of the actuating column of the six-degree-of-freedom dynamic compensation platform, drive the intelligent robotic arm to adjust the attitude of each joint, and calibrate the position of the slide rail. Flexible recovery and recycling preparation module: Receives UAV docking speed data and ship pitch angle data, adjusts the state of the magnetorheological damping device to realize UAV recovery, fixes the UAV through the electronic locking mechanism, initiates the power replenishment process and payload quick change operation, and updates the initial state of the intelligent agent network and data acquisition benchmark.

2. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, The parameter calibration of the information acquisition and intelligent agent networking module includes millimeter-wave array ranging zero-point calibration, visual unit fractal marker coordinate calibration and camera intrinsic parameter calibration, and dual IMU zero-bias calibration. The calibration process is completed by comparing data with a standard attitude reference platform. After calibration, a calibration record is generated and stored. The decentralized navigation intelligent agent network is built through an Ethernet switch. When each intelligent agent node registers, it binds its device identifier, communication address, and data type. During the network self-test, the hardware connection status, data transmission link connectivity, and power supply stability of each node are detected.

3. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the motion prediction and confidence assessment module, the ship-machine cooperative dynamics model is constructed by integrating the dynamic characteristics of the ship and the UAV, including a ship dynamics sub-model, a UAV dynamics sub-model, and a ship-machine coupling mechanism. The ship dynamics sub-model is based on the ship's motion equations, incorporates the effects of wave disturbance and the ship's structural characteristics, inputs the ship's six-degree-of-freedom original motion data, and outputs the trend of the ship's motion attitude change over a short period of time. The UAV dynamics sub-model combines the UAV's mass, moment of inertia, and aerodynamic characteristic parameters, inputs the UAV's real-time flight status data, and outputs the UAV's inertial response characteristic data. The ship-machine coupling mechanism, through a data synchronization interface, fuses the outputs of the ship dynamics sub-model and the UAV dynamics sub-model in real time to construct a unified dynamic mapping relationship. The model output data is used for calculating the fusion confidence and generating motion prediction results.

4. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the motion prediction and confidence assessment module, the mathematical expression of the multidimensional dynamic confidence fusion algorithm is: ,in For the first A navigation agent at any time The dynamic fusion confidence level , , For dynamic adaptive weights, For the first The basic confidence level of a navigation agent itself. For the first The basic confidence level of a navigation agent itself. For the first The number of neighboring effective agents of each navigation agent. For the first The set of neighboring agents of a navigation agent, For the first A navigation agent at any time The pose calculation results, For the first A navigation agent at any time The pose calculation results, It is the minimum value. This is the ship's hull disturbance adaptation coefficient.

5. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the consensus decision and link reconstruction module, after receiving the data, timestamp alignment is first performed. Using the system's reference time axis as a unified reference, linear interpolation is used to correct the timing deviation of data from different navigation agents. A fusion confidence threshold of 0.3 is set. For navigation agents with a fusion confidence lower than the threshold, they are marked as suspected failure nodes and a secondary confirmation process is triggered: three adjacent navigation agents designated by the system verify the data continuity of the suspected node within the most recent preset time period and the consistency between its pose data and the data of the verification node itself; when all three verification nodes report verification failure, the node is determined to be a failure node and is removed; when the navigation fusion architecture is reconstructed, if only a single navigation agent fails, the remaining navigation agents are redistributed according to their respective fusion confidence proportions, and the total weight allocation remains at 1; when multiple navigation agents are disturbed, the top three navigation agents with the highest fusion confidence are extracted first to construct the core navigation link, and the total fusion confidence of the three agents in the core navigation link is not less than 2.

0.

6. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the consensus decision and link reconstruction module, the execution process of the credibility-based weighted voting consensus algorithm is as follows: during the information interaction phase, each navigation agent actively sends its own pose data, fused confidence data and data collection timestamp to all navigation agents in the neighborhood. After receiving neighborhood data, each navigation agent normalizes the data based on the fusion confidence of the other party to obtain the voting weight of the corresponding neighborhood navigation agent. The sum of the voting weights of all neighborhood navigation agents is 1. In the voting phase, each navigation agent votes on the target pose parameters based on its own solution and neighborhood data. The voting result is multiplied by the corresponding weight to obtain the weighted voting value. In the statistical phase, invalid votes with a weighted voting value lower than 0.05 are removed, and the remaining valid weighted voting values ​​are summed to obtain the pose parameter consensus result. Finally, combining the pose consensus results of all valid navigation agents, a unified high-precision fused navigation command is generated through mean fusion, and the average fusion confidence is calculated according to the proportion of fusion confidence of each valid navigation agent.

7. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the dynamic compensation and attitude calibration module, the mathematical expression for the machine-ship linkage adjustment command generation algorithm is: ,in For a moment The set of linkage adjustment commands, The confidence level weighting coefficient is... For all effective navigation agents at time The average fusion confidence, To adjust the gain of the six-degree-of-freedom dynamic compensation platform, For a moment Attitude deviation of the takeoff and landing reference plane, For the attitude adjustment gain of intelligent robotic arms, For a moment The alignment deviation between the centerline of the slide rail and the flight vector of the UAV. To adjust for sea state conditions, For a moment The deviation in the docking speed of the drone This is the time variation of the velocity deviation. For a moment The real-time value of the ship's pitch angle.

8. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the dynamic compensation and attitude calibration module, the six-degree-of-freedom dynamic compensation platform adopts an improved parallel configuration. The upper and lower end plates are symmetrical rigid structures, and the actuating columns are connected to the upper and lower end plates through spherical joints. The actuating columns are arranged symmetrically in pairs to form a spatial triangular configuration, and the extension and retraction of the actuating columns are controlled by a servo drive unit. The intelligent robotic arm adopts a serial configuration, including a rotary joint, a front telescopic mechanism, and an auxiliary fine-tuning joint. It adjusts the slide rail attitude through the coordinated action of multiple joints. During the slide rail calibration process, it receives feedback data from the visual sensor in real time. When a deviation is detected between the slide rail centerline and the UAV flight vector, it immediately recalculates the fine-tuning parameters based on the linkage adjustment command and initiates a secondary attitude correction to ensure that the slide rail and the UAV flight path remain matched.

9. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the flexible recycling and recycling preparation module, the magnetorheological damping device consists of a magnetorheological fluid cavity and a carbon fiber beam segment, arranged in a specific section at the end of the slide rail. The viscosity of the magnetorheological fluid is changed by adjusting the magnetic field strength. The electronically controlled locking mechanism includes an electromagnetic lock and a mechanical positioning pin. After the UAV comes to rest, it is first initially fixed by the electromagnetic lock, and then the mechanical positioning pin is used for secondary limiting. After locking, the locking status is detected by the sensor.

10. The six-degree-of-freedom cooperative control shipborne UAV dynamic deck recovery system according to claim 1, characterized in that, In the flexible recycling and recycling preparation module, the load quick-change interface adopts a modular structure, including a power bus, a data bus, and a mechanical positioning structure. When changing the load, the load tray is pulled out to replace the functional module, and after being pushed back, it is automatically guided to the correct position by the mechanical positioning structure, and the power and data interfaces are connected simultaneously. The entire replacement process does not require a system power outage.

Images