Intelligent cruise and collaborative operation method for multifunctional unmanned ship

Abstract

The invention relates to the field of unmanned ships, and discloses an intelligent cruise and collaborative operation method for a multifunctional unmanned ship, which comprises the following steps: debugging preparation: selecting a functional module according to an operation demand, installing the functional module on an unmanned ship modular platform through a standardized interface, completing initialization and establishing a data transmission link with a cloud; and cruise planning: acquiring environmental data in real time, planning an initial cruise path by using a self-adaptive algorithm in combination with position information acquired by Beidou or GPS navigation in real time, and monitoring environmental changes in real time, dynamically optimizing the path and adjusting a route and a speed to cover a task area during navigation. By collecting environmental data in real time and planning and optimizing a cruise path by using an adaptive algorithm, the unmanned ship can accurately cover a task area, does not depend on a fixed route any more, dynamically adjusts the route and speed according to the actual situation of a water area, quickly reaches a key monitoring point, and reduces unnecessary sailing time.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned ships, and in particular to an intelligent cruising and collaborative operation method of a multifunctional unmanned ship. Background Art

[0002] With the deepening of global ecological governance, water resource management, and environmental information monitoring, unmanned boats, as intelligent, automated mobile surface platforms, are being widely used in various operational scenarios such as water quality monitoring, environmental inspections, emergency search and rescue, and garbage salvage. However, traditional unmanned boat systems mostly adopt a working mode of preset routes and fixed task processes, lacking the ability to respond to real-time environmental changes in the waters, leading to the following prominent problems: (1) The path is highly rigid and cannot be dynamically optimized according to pollution distribution. Traditional geometric paths such as rectangular and spiral planning ignore the heterogeneity of water quality distribution, which can easily lead to repeated patrols in low-pollution areas and missed inspections in high-pollution areas. This wastes energy and makes it difficult to focus on key areas.

[0003] (2) Single task execution, lack of module coordination and dynamic scheduling mechanism. Most unmanned boats can only complete a single function, such as water quality monitoring or garbage collection, and each module works independently of each other. There is a lack of collaborative algorithm support, and it is impossible to perform intelligent task priority assessment and resource reallocation in task-intensive or emergency situations.

[0004] (3) Decision-making relies on the cloud, which lacks real-time performance and edge processing capabilities. High-frequency monitoring data often needs to be uploaded to the cloud for analysis, which has problems such as limited bandwidth and high latency, making it difficult to respond quickly and make intelligent judgments in complex waters.

[0005] (4) Lack of power and status perception capabilities, high risk of mission interruption. Most existing unmanned ships lack complete energy consumption management and return prediction mechanisms. Sudden battery discharge or failure of key modules may lead to mission interruption, data loss, or even equipment damage.

[0006] Therefore, there is an urgent need to build an intelligent unmanned ship system that integrates adaptive path planning, multi-task intelligent scheduling, edge decision analysis and energy status prediction to meet the efficient, reliable and economical operation needs in future multiple water surface operation scenarios. Summary of the Invention

[0007] In response to the shortcomings of the existing technology, the present invention provides an intelligent cruising and collaborative operation method for a multifunctional unmanned boat, which solves the problem that traditional unmanned boats have excessive repeated monitoring in areas with uniform water quality, wasting resources such as electricity and storage space, and are unable to fully monitor in complex polluted areas due to unreasonable routes, thus limiting the role of unmanned boats in water quality monitoring work.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for intelligent cruising and collaborative operation of a multifunctional unmanned vessel, comprising the following steps: S1. Debugging preparation: Select functional modules based on operational requirements, install them on the unmanned vessel modular platform through standardized interfaces, complete initialization, and establish a data transmission link with the cloud; S2. Cruise Planning: Real-time environmental data is collected and combined with real-time location information obtained by Beidou or GPS navigation, an adaptive algorithm is used to plan the initial cruise path. During the voyage, environmental changes are monitored in real time, the path is dynamically optimized, and the route and speed are adjusted to cover the mission area. S3, Dynamic Collaboration: Then, each functional module performs operations in the task area and collects task data. Task priorities are determined according to preset rules and multi-task scheduling algorithms, and tasks are executed or executed in parallel according to task priorities. S4, Processing and Decision-Making: The data is then transmitted to the cloud for analysis using data fusion technology and combined with edge computing for local real-time data processing. S5. Automatic monitoring: Monitor the battery level of the unmanned boat in real time. When the battery level is lower than the preset threshold, plan the return charging route, and continue the operation according to the mission progress after charging is completed. At the same time, monitor the operating status of each functional module.

[0009] Preferably, the functional modules in S1 include a water quality monitoring module, a garbage cleaning module, a rescue response module, a fish and bird identification module, a solar energy module, and a wind energy module.

[0010] Preferably, the step S2 specifically includes the following steps: S201. Environmental Collection: Utilize sensors, sonar, high-definition cameras, and lidar to collect real-time environmental data of the operating waters, while simultaneously constructing a dynamic three-dimensional environmental model and updating it in real time. The sensors include wind speed sensors, water flow sensors, and bioacoustic sensors. The environmental data includes wind and waves, water flow, underwater terrain, obstacle distribution, and underwater biological activity. S202, Path Planning: Obtain the position information of the unmanned vessel based on Beidou or GPS navigation, and use the deep Q network reinforcement learning algorithm in combination with environmental data to perform initial path planning. The position, speed, and environmental information of the unmanned vessel are used as state inputs, discrete actions are defined, and a reward function is designed to guide the algorithm to learn the path strategy. Multi-objective optimization is considered at the same time, and a multi-objective function is constructed. The initial path population is generated using a non-dominated sorting genetic algorithm. The discrete actions include turning, acceleration and deceleration, and the reward function includes goal achievement, obstacle avoidance, path efficiency, and energy consumption. The multi-objective function includes task coverage, energy consumption, operation time, and equipment loss. S203, dynamic optimization: Navigate based on the initial path population. At set intervals or when environmental data changes exceed a threshold, recollect data and update the environmental model. The deep Q-network reinforcement learning algorithm is again used in combination with the real-time environmental status to adjust the path. A non-dominated sorting genetic algorithm is also used to dynamically optimize multiple objectives and select the navigation path. S204, backtracking adjustment: When the planned path is deviated due to a fault, the system automatically backtracks after returning to normal, compares the current position with the original planned path, and replans the path to the unfinished task area by combining real-time environmental data, reinforcement learning, and multi-objective optimization algorithms. The fault includes communication failure and sensor failure.

[0011] Preferably, the step S3 specifically includes the following steps: S301, Mission Collection: Perform operations in the mission area through various functional modules and collect mission data. At the same time, collect external data, including wind speed, water flow, weather, and surrounding ship activities; S302, Priority Determination: Then, the task priority is predicted using a trained machine learning model, and the task priority is determined in combination with preset rules. The machine learning model is trained based on historical task and environmental data and has learned the relationship between tasks and environmental factors. S303, evaluation and allocation: Based on the task priority and the status of the unmanned vessel, resources are allocated using a multi-task scheduling algorithm to execute tasks according to priority or in parallel. The status of the unmanned vessel includes power, storage space, and computing resources; S304, Collaborative Adjustment: During execution, the execution order and collaborative mode are dynamically adjusted according to the progress and resource usage. At the same time, sensors are used to monitor the status of each functional module, predict potential failures, and execute according to the emergency response plan.

[0012] Preferably, the S4 specifically includes the following steps: S401. Data classification: Transmit all data to the cloud and classify them by data type, remove outliers and duplicate data, and perform normalization and feature extraction. S402, real-time processing: using edge computing devices to analyze and process the data after feature extraction to obtain processed data; S403, Cloud Analysis: The processed data is input into a model based on a generative adversarial network. The generator and discriminator are trained adversarially to produce fused data. The inference engine and preset rules are then used to mine the implicit semantic information of the fused data to generate semantic fusion results. Then, big data analysis technology and data fusion algorithms are used to perform deep fusion analysis to obtain analytical data. S404, Feedback Optimization: Combine analysis data and edge computing results to generate decision recommendations to optimize unmanned vessel operations. The decision recommendations include cruise path adjustments, task priority changes, and equipment parameter optimization.

[0013] Preferably, the S5 specifically includes the following steps: S501, automatic monitoring: Use sensors to collect the status of the unmanned vessel in real time, including battery power, charge and discharge parameters, equipment vibration, temperature, and pressure, and perform multi-sensor data fusion; S502, Power Warning: Build a power warning model based on the long short-term memory network algorithm, collect historical operation data for training, input real-time data during operation to predict low battery time, provide early warning, and dynamically adjust the warning threshold based on task progress and environment; S503, Return Plan: Utilize sensors to sense and integrate environmental information, determine path planning conditions based on power consumption and navigation capability, search for a return path, monitor in real time, and adjust the path as needed; S504, Charging Management: After returning home, intelligently adjust charging parameters based on battery status, assess battery health, predict remaining life, and remind you to replace the battery; S505, task recovery: After charging is completed, re-plan based on the real-time environment and task progress and continue to execute the task.

[0014] An intelligent cruising and collaborative operation system for a multifunctional unmanned vessel, comprising: Debugging preparation module: used to select functional modules according to operational requirements, install them into the unmanned vessel modular platform through standardized interfaces, complete initialization, and establish a data transmission link with the cloud; Cruise planning module: This module collects environmental data in real time, combines it with real-time location information obtained by Beidou or GPS navigation, and uses an adaptive algorithm to plan the initial cruise path. It also monitors environmental changes in real time during navigation, dynamically optimizes the path, and adjusts the route and speed to cover the mission area. Dynamic collaboration module: used to perform operations in the task area through various functional modules, collect task data, determine task priorities based on preset rules and multi-task scheduling algorithms, and execute tasks based on task priorities or execute them in parallel; Processing and decision-making module: used to transmit various data to the cloud, analyze them using data fusion technology, and perform local real-time data processing in combination with edge computing; Automatic monitoring module: used to monitor the power level of the unmanned boat in real time. When the power level is lower than the preset threshold, it plans the return charging route and continues the operation according to the mission progress after charging is completed. At the same time, it monitors the operating status of each functional module.

[0015] Preferably, the cruise planning module includes an environment collection unit, a path planning unit, a dynamic optimization unit, and a backtracking adjustment unit. The environment collection unit is used to use sensors, sonar, high-definition cameras and lidar to collect real-time environmental data of the operating waters, and at the same time build a dynamic three-dimensional environmental model and update it in real time. The path planning unit is used to obtain the position information of the unmanned ship through Beidou or GPS navigation, and use the deep Q network reinforcement learning algorithm to perform initial path planning in combination with the environmental data. The position, speed and environmental information of the unmanned ship are used as state inputs to define discrete actions, design reward functions, guide the algorithm to learn path strategies, and consider multi-objective optimization to build a multi-objective reinforcement learning algorithm. The target function is a non-dominated sorting genetic algorithm that generates an initial path population. The dynamic optimization unit is used to navigate according to the initial path population. At set intervals or when environmental data changes exceed a threshold, data is re-collected and the environmental model is updated. The deep Q network reinforcement learning algorithm is used again in combination with the real-time environmental status to adjust the path. At the same time, the non-dominated sorting genetic algorithm is used to dynamically optimize multiple objectives and select a navigation path. The backtracking adjustment unit is used to automatically backtrack after returning to normal when the planned path deviates due to a fault, compare the current position with the original planned path, and replan the path to the unfinished task area by combining real-time environmental data, reinforcement learning, and multi-objective optimization algorithms.

[0016] Preferably, the dynamic collaboration module includes a task collection unit, a priority determination unit, an evaluation and allocation unit, and a collaborative adjustment unit. The task collection unit is used to perform operations in the task area through various functional modules, and collect task data and external data at the same time. The priority determination unit is used to use a trained machine learning model to predict task priority, and at the same time, determine task priority in combination with preset rules. The evaluation and allocation unit is used to allocate resources based on task priority and the status of the unmanned ship using a multi-task scheduling algorithm, and execute tasks according to priority or in parallel. The collaborative adjustment unit is used to dynamically adjust the execution order and collaborative mode according to progress and resource occupancy during execution, and use sensors to monitor the status of each functional module, predict potential faults, and execute according to the emergency response plan.

[0017] The present invention provides a method for intelligent cruising and collaborative operation of a multifunctional unmanned vessel. It has the following beneficial effects: 1. Dynamic route planning accurately covers key waters By introducing reinforcement learning algorithms (such as DQN) and multi-objective optimization (such as NSGA-II), combined with real-time environmental perception to build a three-dimensional water model, breaking the limitations of fixed paths, it can automatically identify high-pollution areas, dynamically avoid obstacles and adjust routes, effectively improving mission coverage efficiency and navigation safety.

[0018] 2. Multi-module collaborative operation, supporting parallel processing and intelligent division of labor It supports plug-and-play modules, including functional units such as water quality monitoring, garbage salvage, and search and rescue response. It can be scheduled in real time according to task complexity, module status, and priority, truly realizing "one ship with multiple functions" and "one task with multiple modes."

[0019] 3. Intelligent determination and scheduling of task priorities The machine learning model is used to automatically determine the priority of the current task in combination with historical tasks and environmental status, and the execution order is dynamically allocated based on resource occupancy, task timeliness, environmental urgency, etc., to optimize resource utilization and ensure that critical tasks are completed first.

[0020] 4. Edge computing + cloud integration to achieve efficient decision-making closed loop The system supports local high-frequency data processing (such as obstacle avoidance control and module self-test), and uploads the integrated key data to the cloud for large-scale pattern recognition, strategy generation and task optimization, realizing the dual decision-making path of "fast response + deep insight".

[0021] 5. Have the ability to backtrack after task interruption to improve task integrity After a communication anomaly or system failure occurs, the system can automatically backtrack and re-compare the current and original path statuses in the recovery state, determine the unfinished task areas and accurately re-measure, effectively solving the breakpoint problems left over from traditional systems.

[0022] 6. Battery prediction and intelligent return to ensure endurance safety An LSTM network is introduced to predict low battery times, and the return strategy is dynamically adjusted based on mission progress and environmental conditions. Healthy charging management and battery life prediction are supported to ensure uninterrupted navigation missions and long-term stable equipment operation.

[0023] 7. Multi-source environmental perception improves emergency response capabilities The system supports access to multi-dimensional sensors such as wind speed, water flow, bioacoustics, and lidar. It has the ability to construct real-time maps of the water surface, underwater, and boundary environments, and can quickly avoid and reconstruct paths for obstacles, sudden pollution sources, or weather changes.

[0024] 8. Deep data fusion improves monitoring quality and strategy recommendations Generative adversarial networks (GANs) are used to fuse data streams from different modules, combined with semantic reasoning and rule engines to automatically identify potential pollution trends and ecological risks, assist decision-making systems in adjusting routes and task configurations, and promote the construction of an intelligent water management system.

[0025] 9. Modular architecture supports flexible expansion of multi-tasking scenarios The unmanned boat platform is designed with standardized interfaces, which can flexibly switch functional components according to the mission scenario. It also has good upgrade compatibility and is suitable for various application scenarios such as ecological inspections, scientific research investigations, and urban water management. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a method flow chart of the intelligent cruising and collaborative operation method of a multifunctional unmanned vessel proposed by the present invention; Figure 2 This is a cruise planning flow chart of the intelligent cruise and collaborative operation method of a multifunctional unmanned ship proposed in the present invention; Figure 3 A dynamic collaborative flow chart of the intelligent cruising and collaborative operation method of a multifunctional unmanned vessel proposed in the present invention; Figure 4 This is a decision-making flow chart for the intelligent cruising and collaborative operation method of a multifunctional unmanned vessel proposed in the present invention; Figure 5 This is a flow chart of automatic monitoring of the intelligent cruising and collaborative operation method of a multifunctional unmanned vessel proposed by the present invention; Figure 6 This is a system architecture diagram of the intelligent cruising and collaborative operation system of a multifunctional unmanned vessel proposed in the present invention; Figure 7 This is a diagram of the cruise planning module architecture of the intelligent cruise and collaborative operation system of a multifunctional unmanned vessel proposed in the present invention; Figure 8 This is a dynamic collaborative module architecture diagram of the intelligent cruising and collaborative operation system of a multifunctional unmanned ship proposed by the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] Example 1: Please see the attached Figure 1 The embodiment of the present invention provides a method for intelligent cruising and collaborative operation of a multifunctional unmanned vessel, comprising the following steps: S1. Debugging preparation: Select functional modules based on operational requirements, install them on the unmanned boat modular platform through standardized interfaces, complete initialization, and establish a data transmission link with the cloud; the functional modules in S1 include water quality monitoring module, garbage cleaning module, rescue response module, fish and bird identification module, solar module, and wind power module.

[0029] S2. Cruise Planning: Real-time environmental data is collected and combined with real-time location information obtained by Beidou or GPS navigation, an adaptive algorithm is used to plan the initial cruise path. During the voyage, environmental changes are monitored in real time, the path is dynamically optimized, and the route and speed are adjusted to cover the mission area. S3, Dynamic Collaboration: Then, each functional module performs operations in the task area and collects task data. Task priorities are determined according to preset rules and multi-task scheduling algorithms, and tasks are executed or executed in parallel according to task priorities. S4, Processing and Decision-Making: The data is then transmitted to the cloud for analysis using data fusion technology and combined with edge computing for local real-time data processing. S5. Automatic monitoring: Monitor the battery level of the unmanned boat in real time. When the battery level is lower than the preset threshold, plan the return charging route, and continue the operation according to the mission progress after charging is completed. At the same time, monitor the operating status of each functional module.

[0030] Specifically, water quality monitoring, garbage collection, solar energy, and wind energy modules were selected based on the operational needs of these systems. These modules were then installed on the unmanned vessel's modular platform using standardized interfaces. Initialization was completed by calibrating sensors and inspecting the mechanical structure, and establishing a stable data transmission link with the cloud. The standardized interfaces ensured stable module connections, initialization ensured normal equipment operation, and the data link laid the foundation for subsequent data transmission, providing fundamental support for operations.

[0031] After the unmanned vessel sets sail, multiple sensors collaboratively collect environmental data, such as wind speed and underwater topography, to construct and update a dynamic three-dimensional environmental model in real time. Combined with navigational location information, a deep Q-network reinforcement learning algorithm is used to plan an initial path. This algorithm considers multiple objectives and constructs a function to generate a population of paths. If timing or environmental changes exceed a threshold during navigation, data is re-collected to optimize the path. The route and speed are adjusted accordingly to ensure precise coverage of the mission area, avoid over-monitoring in areas of uniform water quality, and rationally plan routes in complex, polluted areas.

[0032] The water quality monitoring and garbage collection modules operate simultaneously, collecting tasks and external data. A trained machine learning model, combined with pre-set rules, prioritizes tasks, such as prioritizing monitoring when water quality is abnormal. A multi-task scheduling algorithm allocates resources based on priority and the availability of unmanned vessels, enabling parallel or sequential execution of tasks. During execution, the order and coordination methods are dynamically adjusted, module status is monitored, and potential failures are addressed according to pre-planned procedures to ensure smooth operation.

[0033] The unmanned vessel transmits collected data to the cloud for classification, denoising, and feature extraction. Edge computing devices process real-time critical data, such as obstacle avoidance. The cloud uses a generative adversarial network to fuse the data and mine semantic information. After in-depth analysis, it combines edge computing results to generate decision-making recommendations, such as adjusting cruise paths, changing mission priorities, and optimizing equipment parameters. This feedback is used to optimize the unmanned vessel's operations.

[0034] Sensors collect and integrate real-time data on the drone's battery level and equipment status. A model based on a long-short-term memory network algorithm predicts low battery times and issues warnings, adjusting thresholds based on the mission and environment. When the battery is low, the system plans a return route, intelligently charges the drone upon return, and assesses battery health. Once charging is complete, the system replans the route based on real-time conditions and resumes unfinished tasks, ensuring operational continuity.

[0035] By collecting environmental data in real time during the cruise planning phase and applying advanced algorithms to dynamically optimize routes, the unmanned vessel can rationally plan routes based on the actual conditions of the water area, avoiding excessive monitoring in areas with uniform water quality and reducing the waste of power and storage space. During the dynamic collaboration phase, by accurately determining task priorities and rationally allocating resources, the unmanned vessel can prioritize key tasks in complex polluted areas, improving monitoring and cleanup efficiency. During the decision-making phase, a combination of cloud and edge computing is used to deeply analyze data and optimize operational decisions based on the results. During the automatic monitoring phase, power and equipment status are monitored in real time to ensure the stable operation of the unmanned vessel and the continuity of its mission. This enables intelligent cruise and efficient collaborative operations for unmanned vessels in diverse water environments, resolving the issues faced by traditional unmanned vessels, such as excessive and repeated monitoring in areas with uniform water quality, which wastes resources such as power and storage space, and the difficulty of adequately monitoring complex polluted areas due to irrational routing, which limits the role of unmanned vessels in water quality monitoring.

[0036] Please see the attached Figure 2 , S2 specifically includes the following steps: S201, Environmental Collection: Utilize sensors, sonar, high-definition cameras, and lidar to collect real-time environmental data of the operating waters, while simultaneously constructing a dynamic three-dimensional environmental model and updating it in real time. Sensors include wind speed sensors, water flow sensors, and bioacoustic sensors. Environmental data includes wind and waves, water flow, underwater terrain, obstacle distribution, and underwater biological activity. S202, Path Planning: Obtain the position information of the unmanned vessel based on Beidou or GPS navigation, and use the deep Q network reinforcement learning algorithm in combination with environmental data for initial path planning. The position, speed, and environmental information of the unmanned vessel are used as state inputs to define discrete actions, design a reward function, and guide the algorithm to learn the path strategy. At the same time, multi-objective optimization is considered, a multi-objective function is constructed, and a non-dominated sorting genetic algorithm is used to generate the initial path population. Discrete actions include turning, acceleration and deceleration. The reward function includes goal achievement, obstacle avoidance, path efficiency, and energy consumption. The multi-objective function includes task coverage, energy consumption, operation time, and equipment loss. S203, dynamic optimization: Navigate based on the initial path population. At set intervals or when environmental data changes exceed a threshold, recollect data and update the environmental model. The deep Q-network reinforcement learning algorithm is again used in combination with the real-time environmental status to adjust the path. A non-dominated sorting genetic algorithm is also used to dynamically optimize multiple objectives and select the navigation path. S204, backtracking adjustment: When the planned path is deviated due to a fault, the system will automatically backtrack after returning to normal, compare the current position with the original planned path, and re-plan the path to the unfinished task area by combining real-time environmental data, reinforcement learning and multi-objective optimization algorithms. Faults include communication failures and sensor failures.

[0037] Specifically, once the unmanned boat enters the operating area, wind speed sensors, water flow sensors, bioacoustic sensors, sonar, high-definition cameras, and lidar work in tandem. The wind speed sensor monitors the wind speed and direction of the lake surface in real time, the water flow sensor captures the speed and direction of the water flow, the sonar detects the underwater topography, the high-definition camera and lidar monitor the water surface and surrounding environment, and the bioacoustic sensor collects sound signals generated by underwater biological activity. These devices collect data on wind, waves, water flow, underwater topography, obstacle distribution, and underwater biological activity, and use it to construct a dynamic three-dimensional environmental model. This model updates in real time as the environment changes, providing accurate environmental information for subsequent path planning.

[0038] Based on the Beidou or GPS navigation system, the unmanned boat's real-time location information is obtained. Combined with environmental data, a deep Q-network reinforcement learning algorithm is used for initial path planning. The unmanned boat's position, speed, and current environmental information are used as state inputs to define discrete actions such as steering, acceleration, and deceleration. When designing the reward function, factors such as goal achievement, obstacle avoidance, path efficiency, and energy consumption are fully considered. If the unmanned boat successfully reaches a preset water quality monitoring point or garbage collection area, a goal achievement reward is given; during the obstacle avoidance process, successful obstacle avoidance rewards are awarded. The more efficient the path planning, the higher the path efficiency reward. At the same time, energy consumption is minimized to conserve energy. A multi-objective function is constructed, encompassing task coverage, energy consumption, operation time, and equipment loss. A non-dominated sorting genetic algorithm is used to generate the initial path population, ensuring that path planning comprehensively considers the balance of multiple objectives.

[0039] The unmanned vessel begins navigating along an initial population of paths. At regular intervals (e.g., 10 minutes) or when environmental data changes exceed a preset threshold (e.g., sudden changes in wind speed, the appearance of a new obstacle), it recollects data and updates the environmental model. The deep Q-network reinforcement learning algorithm, combined with real-time environmental conditions, adjusts the path, while a non-dominated sorting genetic algorithm dynamically optimizes multiple objectives. When adjusting the path, the optimal navigation path is selected, taking into account current environmental factors, mission progress, and energy consumption, ensuring the unmanned vessel consistently completes its mission efficiently.

[0040] During navigation, if a communication or sensor failure causes the unmanned vessel to deviate from the planned path, it will automatically backtrack after the fault is restored. By comparing its current position with the original planned path and combining it with real-time environmental data, the unmanned vessel uses reinforcement learning and multi-objective optimization algorithms to re-plan a path to the unfinished mission area, ensuring mission continuity and integrity.

[0041] Please see the attached Figure 3 , S3 specifically includes the following steps: S301, Mission Collection: Perform operations in the mission area through various functional modules and collect mission data. At the same time, collect external data, including wind speed, water flow, weather, and surrounding ship activities; S302, Priority Determination: Then use the trained machine learning model to predict task priority and combine it with preset rules to determine task priority. The machine learning model is trained based on historical task and environmental data and has learned the relationship between tasks and environmental factors. S303, evaluation and allocation: Based on the priority of tasks and the status of the unmanned vessel, a multi-task scheduling algorithm is used to allocate resources, and tasks are executed according to priority or in parallel. The status of the unmanned vessel includes power, storage space, and computing resources; S304, Collaborative Adjustment: During execution, the execution order and collaborative mode are dynamically adjusted according to the progress and resource usage. At the same time, sensors are used to monitor the status of each functional module, predict potential failures, and execute according to the emergency response plan.

[0042] Specifically, each functional module performs operations in the task area and collects task data, while also collecting external data to provide comprehensive information for subsequent task priority determination.

[0043] Task priorities are predicted using a pre-trained machine learning model. This model, trained on historical task and environmental data, has learned the relationships between tasks and environmental factors. Prioritization is determined based on pre-set rules (e.g., prioritizing water quality monitoring tasks when water quality monitoring data is abnormal, or prioritizing garbage collection tasks when garbage accumulation areas expand).

[0044] A multi-task scheduling algorithm allocates resources based on task priority, taking into account the boat's power, storage, and computing resources. High-priority tasks are prioritized with sufficient resources. If tasks can be executed in parallel and resources allow, resources are allocated appropriately to ensure efficient execution of multiple tasks. For example, if the water quality monitoring task has a high priority and sufficient power is available, energy supply to the water quality monitoring module is prioritized, while computing resources are allocated appropriately for data analysis.

[0045] During mission execution, the execution order and coordination methods are dynamically adjusted based on task progress and resource usage. Sensors monitor the operating status of each functional module and predict potential failures. If the garbage collection module faces the risk of mechanical failure, the task schedule is adjusted in advance, prioritizing water quality monitoring tasks and promptly initiating emergency response plans to ensure safe and stable unmanned boat operations.

[0046] Please see the attached Figure 4 , S4 specifically includes the following steps: S401. Data classification: Transmit all data to the cloud and classify them by data type, remove outliers and duplicate data, and perform normalization and feature extraction. S402, real-time processing: using edge computing devices to analyze and process the data after feature extraction to obtain processed data; S403, Cloud Analysis: The processed data is input into a model based on a generative adversarial network. The generator and discriminator are trained adversarially to produce fused data. The inference engine and preset rules are then used to mine the implicit semantic information of the fused data to generate semantic fusion results. Then, big data analysis technology and data fusion algorithms are used to perform deep fusion analysis to obtain analytical data. S404, Feedback Optimization: Combine analytical data and edge computing results to generate decision recommendations to optimize unmanned vessel operations. Decision recommendations include cruise path adjustments, task priority changes, and equipment parameter optimization.

[0047] Specifically, the data collected by the unmanned boats is transmitted to the cloud and categorized by data type, such as water quality monitoring data, garbage collection data, and environmental data. Outliers and duplicate data are removed, and the data is normalized to make different types of data comparable. Key features are then extracted for subsequent analysis.

[0048] Edge computing devices are used to analyze and process data after feature extraction, and to quickly process some data with high real-time requirements, such as emergency obstacle avoidance information and equipment failure warnings, to reduce cloud computing pressure and ensure that unmanned ships can respond to emergencies in a timely manner.

[0049] The processed data is fed into a generative adversarial network-based model. The generator and discriminator are trained adversarially to produce fused data, exploring potential connections between the data. An inference engine and pre-set rules are then used to mine the implicit semantic information within the fused data, generating a semantically fused result. Big data analysis techniques and data fusion algorithms are then used to perform in-depth fusion analysis of the data, yielding more comprehensive and accurate analytical data.

[0050] Combining analytical data with edge computing results, it generates decision recommendations, such as adjusting the cruise path to enable the unmanned vessel to complete its mission more efficiently, changing mission priorities to address emergencies, and optimizing equipment parameters to improve performance. These recommendations are then fed back to the unmanned vessel to optimize its operations.

[0051] Please see the attached Figure 5 , S5 specifically includes the following steps: S501, automatic monitoring: Use sensors to collect the status of the unmanned vessel in real time, including battery power, charge and discharge parameters, equipment vibration, temperature, and pressure, and perform multi-sensor data fusion; S502, Power Warning: Build a power warning model based on the long short-term memory network algorithm, collect historical operation data for training, input real-time data during operation to predict low battery time, provide early warning, and dynamically adjust the warning threshold based on task progress and environment; S503, Return Plan: Utilize sensors to sense and integrate environmental information, determine path planning conditions based on power consumption and navigation capability, search for a return path, monitor in real time, and adjust the path as needed; S504, Charging Management: After returning home, intelligently adjust charging parameters based on battery status, assess battery health, predict remaining life, and remind you to replace the battery; S505, task recovery: After charging is completed, re-plan based on the real-time environment and task progress and continue to execute the task.

[0052] Specifically, sensors are used to collect real-time information on the battery power, charging and discharging parameters, and equipment vibration, temperature, pressure, and other status information of the unmanned boat, and multi-sensor data is fused to fully understand the operating status of the unmanned boat.

[0053] A battery warning model based on a long-short-term memory network algorithm was constructed, using historical operational data from unmanned vessels for training. During operation, real-time data input was used to predict low battery times and issue early warnings. Warning thresholds were dynamically adjusted based on mission progress and environmental factors. For example, in severe weather or complex mission environments, early warnings were issued to ensure the unmanned vessel had sufficient battery power to return.

[0054] When the battery level drops below a preset threshold, sensors sense and integrate environmental information, determining a path planning condition based on battery consumption and navigation capability, and searching for a return route. During the return journey, environmental changes are monitored in real time and the route is adjusted as needed to ensure a safe and efficient return.

[0055] After returning home, the unmanned vessel intelligently adjusts charging parameters based on battery status, adopting the appropriate charging mode to protect the battery and improve charging efficiency. It also assesses battery health and predicts remaining life, alerting staff to replace batteries when they are nearing their life limit.

[0056] After charging is completed, the system will re-plan based on the real-time environment and task progress, and continue to execute unfinished tasks to ensure continuous and efficient progress of tasks.

[0057] Example 2: Please see the attached Figure 6 , an intelligent cruising and collaborative operation system for a multifunctional unmanned vessel, including: Debugging preparation module: used to select functional modules according to operational requirements, install them into the unmanned vessel modular platform through standardized interfaces, complete initialization, and establish a data transmission link with the cloud; Cruise planning module: This module collects environmental data in real time, combines it with real-time location information obtained by Beidou or GPS navigation, and uses an adaptive algorithm to plan the initial cruise path. It also monitors environmental changes in real time during navigation, dynamically optimizes the path, and adjusts the route and speed to cover the mission area. Dynamic collaboration module: used to perform operations in the task area through various functional modules, collect task data, determine task priorities based on preset rules and multi-task scheduling algorithms, and execute tasks based on task priorities or execute them in parallel; Processing and decision-making module: used to transmit various data to the cloud, analyze them using data fusion technology, and perform local real-time data processing in combination with edge computing; Automatic monitoring module: used to monitor the power level of the unmanned boat in real time. When the power level is lower than the preset threshold, it plans the return charging route and continues the operation according to the mission progress after charging is completed. At the same time, it monitors the operating status of each functional module.

[0058] Specifically, by selecting functional modules according to operational requirements, installing them on the unmanned ship modular platform through standardized interfaces, completing initialization and establishing a data transmission link with the cloud, it is ensured that the unmanned ship can be equipped with functional modules that meet specific tasks, achieving plug-and-play and greatly improving the applicability of the unmanned ship. The standardized interface makes the installation of each module as convenient as splicing building blocks, reducing the time cost of installation and debugging. The initialization operation ensures that each functional module is in the best working condition before the operation, improving the stability of the equipment operation. The data transmission link established with the cloud provides a channel for the real-time transmission and analysis of subsequent data, so that the unmanned ship can obtain computing resources and historical data support from the cloud at any time during the operation, laying the foundation for intelligent decision-making.

[0059] By collecting environmental data in real time, combined with real-time location information obtained by Beidou or GPS navigation, an adaptive algorithm is used to plan the initial cruise path. Environmental changes are monitored in real time during navigation, and the path is dynamically optimized, with the route and speed adjusted to cover the mission area. This allows the unmanned vessel to plan the optimal navigation path based on real-time environmental conditions. Environmental data collected collaboratively by multiple sensors, such as wind speed, water currents, underwater terrain, and obstacle distribution, provides comprehensive information for path planning. The adaptive algorithm, combined with navigation position information, fully considers the unmanned vessel's current position and mission objectives to plan a reasonable initial path. During navigation, real-time monitoring of environmental changes and dynamic path optimization ensure that the unmanned vessel can avoid obstacles, adapt to complex water currents and weather conditions, and efficiently cover the mission area, improving operational efficiency while ensuring the unmanned vessel's navigation safety.

[0060] Through various functional modules operating in the task area and collecting task data, the task priority is determined according to preset rules and multi-task scheduling algorithms, and tasks are executed according to the task priority or in parallel, thereby realizing efficient collaborative processing of multiple tasks. Different functional modules perform their respective duties in the same task area, such as the water quality monitoring module collecting water quality data, the garbage cleaning module salvaging garbage, etc. At the same time, the collected task data and external environment data provide a basis for determining the task priority. The preset rules and multi-task scheduling algorithms can reasonably allocate the computing resources, energy and storage space of the unmanned ship according to factors such as the urgency of the task, resource requirements and environmental changes, ensuring that important tasks are executed first, and parallel tasks can be carried out at the same time, avoiding resource conflicts and improving the overall operation capability of the unmanned ship.

[0061] By transmitting various data to the cloud, analyzing them using data fusion technology, and combining them with edge computing for local real-time data processing, in-depth data analysis and rapid response are achieved. Edge computing devices quickly process data with high real-time requirements locally, such as obstacle avoidance decisions and equipment failure warnings, reducing the computing burden on the cloud and ensuring that unmanned ships can respond to emergencies in a timely manner. The data transmitted to the cloud is analyzed through data fusion technology to discover the potential connections and patterns between different types of data. For example, by combining water quality monitoring data with environmental data, the possible location and propagation path of the pollution source can be analyzed, thereby optimizing the unmanned ship's operating strategy.

[0062] By monitoring the power of the unmanned ship in real time, when the power is lower than the preset threshold, the return charging route is planned, and the operation is continued according to the mission progress after charging is completed. At the same time, the operating status of each functional module is monitored, thereby ensuring the continuity and stability of the unmanned ship operation. The power monitoring function enables the unmanned ship to plan the return charging in advance to avoid mission interruption or equipment loss due to power exhaustion. During the return process, the optimal charging route is planned according to real-time environmental information, which improves energy utilization efficiency. After charging is completed, the operation is re-planned according to the mission progress to ensure that the mission can be smoothly carried out. Real-time monitoring of the operating status of each functional module can timely discover potential faults and take measures to repair or adjust them in advance, reducing the impact of equipment failures on operations and ensuring the stability of the entire operation process.

[0063] Through the coordinated operation of various modules, an intelligent and efficient unmanned boat operation system has been built, realizing the intelligent cruising and collaborative operation of multifunctional unmanned boats in complex water environments, improving the quality and effectiveness of water operations, and providing strong technical support for aquatic ecological protection, resource development and other fields.

[0064] Please see the attached Figure 7 The cruise planning module includes an environment acquisition unit, a path planning unit, a dynamic optimization unit, and a backtracking adjustment unit. The environment acquisition unit is used to use sensors, sonar, high-definition cameras and lidar to collect real-time environmental data of the operating waters, and at the same time build a dynamic three-dimensional environmental model and update it in real time. The path planning unit is used to obtain the position information of the unmanned ship through Beidou or GPS navigation, and use the deep Q network reinforcement learning algorithm to perform initial path planning in combination with the environmental data. The position, speed and environmental information of the unmanned ship are used as state inputs, discrete actions are defined, and reward functions are designed to guide the algorithm to learn path strategies. At the same time, multi-objective optimization is considered and a multi-objective function is constructed. The non-dominated sorting genetic algorithm is used to generate the initial path population. The dynamic optimization unit is used to navigate according to the initial path population. Every set time or when the environmental data changes exceed the threshold, the data is re-collected and the environmental model is updated. The deep Q network reinforcement learning algorithm is used again in combination with the real-time environmental status to adjust the path. At the same time, the non-dominated sorting genetic algorithm is used to dynamically optimize multiple objectives and select the navigation path. The backtracking adjustment unit is used to automatically backtrack after returning to normal when the planned path deviates due to a fault, compare the current position with the original planned path, and re-plan the path to the unfinished task area by combining real-time environmental data, reinforcement learning and multi-objective optimization algorithms.

[0065] Specifically, the environmental acquisition unit's wind speed sensors, water flow sensors, bioacoustic sensors, sonar, high-definition cameras, and lidar work in tandem. During monitoring, the wind speed sensor captures real-time changes in wind speed and direction on the lake surface, while the water flow sensor accurately measures the speed and direction of the water flow, providing essential data for the unmanned vessel's navigation. The sonar continuously scans the underwater terrain, while the high-definition cameras and lidar monitor the water surface and surrounding environment, promptly detecting obstacles. The bioacoustic sensor collects information on underwater biological activity. This data is used to construct a dynamic three-dimensional environmental model. For example, if a sudden strong wind strikes the lake, the model can be rapidly updated, providing an accurate basis for path planning.

[0066] The path planning unit uses Beidou or GPS navigation to obtain the UAV's position information. Combined with environmental data, it uses a deep Q-network reinforcement learning algorithm to plan a path. The UAV's position, speed, and environmental information are used as state inputs to define discrete actions such as steering, acceleration, and deceleration. A reward function is designed, rewarding successful obstacle avoidance and rapid arrival at monitoring points to guide algorithm learning. A multi-objective function is constructed, taking into account multiple objectives such as mission coverage and energy consumption. A non-dominated sorting genetic algorithm is used to generate an initial path population, ensuring that the planned path balances mission completion with resource conservation.

[0067] During navigation, the dynamic optimization unit triggers the environmental collection unit to re-collect data and update the environmental model at regular intervals or when environmental data changes exceed a threshold. The deep Q-network reinforcement learning algorithm, combined with real-time environmental conditions, adjusts the route, and a non-dominated sorting genetic algorithm dynamically optimizes multiple objectives. If a new garbage collection area is encountered, the route is replanned to clean it up, improving operational efficiency.

[0068] If the unmanned vessel deviates from its planned path due to a communication or sensor failure, the backtracking adjustment unit activates after the fault is resolved. It compares its current location with the original planned path and, combining real-time environmental data with reinforcement learning and multi-objective optimization algorithms, replans a route to the unfinished mission area. If a fault causes the vessel to miss a water quality monitoring point, the vessel will re-route to the monitoring point after recovery to ensure mission integrity.

[0069] Please see the attached Figure 8The dynamic collaboration module includes a task collection unit, a priority determination unit, an evaluation and allocation unit, and a collaborative adjustment unit. The task collection unit is used to perform operations in the task area through various functional modules, collect task data, and collect external data at the same time. The priority determination unit is used to use the trained machine learning model to predict the task priority, and at the same time, combine the preset rules to determine the task priority. The evaluation and allocation unit is used to allocate resources based on the task priority and the status of the unmanned ship using a multi-task scheduling algorithm, execute tasks according to priority or in parallel, and the collaborative adjustment unit is used to dynamically adjust the execution order and collaborative mode according to the progress and resource occupancy during execution. At the same time, it uses sensors to monitor the status of each functional module, predict potential failures, and execute according to the emergency response plan.

[0070] Specifically, the mission collection unit enables the water quality monitoring module and the garbage removal module to operate simultaneously within the mission area. The water quality monitoring module collects water samples at set intervals to test indicators such as pH and dissolved oxygen; the garbage removal module activates a collection device to salvage garbage. It also collects external data such as wind speed, current, weather, and surrounding vessel activity. In the event of heavy rainfall, current speed and water quality may change. Timely collection of this data supports subsequent decision-making.

[0071] The priority determination unit uses a trained machine learning model based on historical tasks and environmental data to predict task priorities. For example, based on past experience, when rapid water flow affects waste removal efficiency and water quality data is abnormal, the model predicts a higher priority for water quality monitoring tasks. Combined with pre-set rules (e.g., prioritizing monitoring when water pollution is severe), task priorities are determined comprehensively.

[0072] The assessment and allocation unit allocates resources using a multi-task scheduling algorithm based on task priority, combined with the UAV's battery capacity, storage space, and computing resources. When the water quality monitoring task has a high priority and sufficient battery capacity, computing resources are prioritized for data analysis in the water quality monitoring module, while the garbage collection module's operating hours are appropriately scheduled to ensure efficient resource utilization.

[0073] During task execution, the coordination and adjustment unit dynamically adjusts the execution order and coordination methods based on task progress and resource usage. Sensors monitor the status of each functional module. If the garbage collection module faces the risk of mechanical failure, the task schedule is adjusted in advance, prioritizing water quality monitoring tasks. Equipment is then repaired or replaced according to the emergency response plan to ensure stable operations.

[0074] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent cruising and collaborative operation of a multifunctional unmanned vessel, characterized in that: The following steps are involved: S1. Debugging preparation: Select functional modules based on operational requirements, install them on the unmanned vessel modular platform through standardized interfaces, complete initialization, and establish a data transmission link with the cloud; S2. Cruise Planning: Real-time environmental data is collected and combined with real-time location information obtained by Beidou or GPS navigation, an adaptive algorithm is used to plan the initial cruise path. During the voyage, environmental changes are monitored in real time, the path is dynamically optimized, and the route and speed are adjusted to cover the mission area. S3, Dynamic Collaboration: Then, each functional module performs operations in the task area and collects task data. Task priorities are determined according to preset rules and multi-task scheduling algorithms, and tasks are executed or executed in parallel according to task priorities. S4, Processing and Decision-Making: The data is then transmitted to the cloud for analysis using data fusion technology and combined with edge computing for local real-time data processing. S5. Automatic monitoring: Monitor the battery level of the unmanned boat in real time. When the battery level is lower than the preset threshold, plan the return charging route, and continue the operation according to the mission progress after charging is completed. At the same time, monitor the operating status of each functional module.

2. The intelligent cruising and collaborative operation method of a multifunctional unmanned vessel according to claim 1, characterized in that: The functional modules in S1 include a water quality monitoring module, a garbage cleaning module, a rescue response module, a fish and bird identification module, a solar energy module, and a wind energy module.

3. The intelligent cruising and collaborative operation method of a multifunctional unmanned vessel according to claim 1, characterized in that: The S2 specifically includes the following steps: S201. Environmental Collection: Utilize sensors, sonar, high-definition cameras, and lidar to collect real-time environmental data of the operating waters, while simultaneously constructing a dynamic three-dimensional environmental model and updating it in real time. The sensors include wind speed sensors, water flow sensors, and bioacoustic sensors. The environmental data includes wind and waves, water flow, underwater terrain, obstacle distribution, and underwater biological activity. S202, Path Planning: Obtain the position information of the unmanned vessel based on Beidou or GPS navigation, and use the deep Q network reinforcement learning algorithm in combination with environmental data to perform initial path planning. The position, speed, and environmental information of the unmanned vessel are used as state inputs, discrete actions are defined, and a reward function is designed to guide the algorithm to learn the path strategy. Multi-objective optimization is considered at the same time, and a multi-objective function is constructed. The initial path population is generated using a non-dominated sorting genetic algorithm. The discrete actions include turning, acceleration and deceleration, and the reward function includes goal achievement, obstacle avoidance, path efficiency, and energy consumption. The multi-objective function includes task coverage, energy consumption, operation time, and equipment loss. S203, dynamic optimization: Navigate based on the initial path population. At set intervals or when environmental data changes exceed a threshold, recollect data and update the environmental model. The deep Q-network reinforcement learning algorithm is again used in combination with the real-time environmental status to adjust the path. A non-dominated sorting genetic algorithm is also used to dynamically optimize multiple objectives and select the navigation path. S204, backtracking adjustment: When the planned path is deviated due to a fault, the system automatically backtracks after returning to normal, compares the current position with the original planned path, and replans the path to the unfinished task area by combining real-time environmental data, reinforcement learning, and multi-objective optimization algorithms. The fault includes communication failure and sensor failure.

4. The intelligent cruising and collaborative operation method of a multifunctional unmanned vessel according to claim 1, characterized in that: The S3 specifically includes the following steps: S301, Mission Collection: Perform operations in the mission area through various functional modules and collect mission data. At the same time, collect external data, including wind speed, water flow, weather, and surrounding ship activities; S302, Priority Determination: Then, the task priority is predicted using a trained machine learning model, and the task priority is determined in combination with preset rules. The machine learning model is trained based on historical task and environmental data and has learned the relationship between tasks and environmental factors. S303, evaluation and allocation: Based on the task priority and the status of the unmanned vessel, resources are allocated using a multi-task scheduling algorithm to execute tasks according to priority or in parallel. The status of the unmanned vessel includes power, storage space, and computing resources; S304, Collaborative Adjustment: During execution, the execution order and collaborative mode are dynamically adjusted according to the progress and resource usage. At the same time, sensors are used to monitor the status of each functional module, predict potential failures, and execute according to the emergency response plan.

5. The intelligent cruising and collaborative operation method of a multifunctional unmanned vessel according to claim 1, characterized in that: The S4 specifically includes the following steps: S401. Data classification: Transmit all data to the cloud and classify them by data type, remove outliers and duplicate data, and perform normalization and feature extraction. S402, real-time processing: using edge computing devices to analyze and process the data after feature extraction to obtain processed data; S403, Cloud Analysis: The processed data is input into a model based on a generative adversarial network. The generator and discriminator are trained adversarially to produce fused data. The inference engine and preset rules are then used to mine the implicit semantic information of the fused data to generate semantic fusion results. Then, big data analysis technology and data fusion algorithms are used to perform deep fusion analysis to obtain analytical data. S404, Feedback Optimization: Combine analysis data and edge computing results to generate decision recommendations to optimize unmanned vessel operations. The decision recommendations include cruise path adjustments, task priority changes, and equipment parameter optimization.

6. The intelligent cruising and collaborative operation method of a multifunctional unmanned vessel according to claim 1, characterized in that: The S5 specifically includes the following steps: S501, automatic monitoring: Use sensors to collect the status of the unmanned vessel in real time, including battery power, charge and discharge parameters, equipment vibration, temperature, and pressure, and perform multi-sensor data fusion; S502, Power Warning: Build a power warning model based on the long short-term memory network algorithm, collect historical operation data for training, input real-time data during operation to predict low battery time, provide early warning, and dynamically adjust the warning threshold based on task progress and environment; S503, Return Plan: Utilize sensors to sense and integrate environmental information, determine path planning conditions based on power consumption and navigation capability, search for a return path, monitor in real time, and adjust the path as needed; S504, Charging Management: After returning home, intelligently adjust charging parameters based on battery status, assess battery health, predict remaining life, and remind you to replace the battery; S505, task recovery: After charging is completed, re-planning is performed based on the real-time environment and task progress, and the task is continued.

7. An intelligent cruising and collaborative operation system for a multifunctional unmanned vessel, characterized by: An intelligent cruising and collaborative operation method for a multifunctional unmanned ship according to any one of claims 1 to 6, comprising: Debugging preparation module: used to select functional modules according to operational requirements, install them into the unmanned vessel modular platform through standardized interfaces, complete initialization, and establish a data transmission link with the cloud; Cruise planning module: This module collects environmental data in real time, combines it with real-time location information obtained by Beidou or GPS navigation, and uses an adaptive algorithm to plan the initial cruise path. It also monitors environmental changes in real time during navigation, dynamically optimizes the path, and adjusts the route and speed to cover the mission area. Dynamic collaboration module: used to perform operations in the task area through various functional modules, collect task data, determine task priorities based on preset rules and multi-task scheduling algorithms, and execute tasks based on task priorities or execute them in parallel; Processing and decision-making module: used to transmit various data to the cloud, analyze them using data fusion technology, and perform local real-time data processing in combination with edge computing; Automatic monitoring module: used to monitor the power level of the unmanned boat in real time. When the power level is lower than the preset threshold, it plans the return charging route and continues the operation according to the mission progress after charging is completed. At the same time, it monitors the operating status of each functional module.

8. The intelligent cruising and collaborative operation system for a multifunctional unmanned vessel according to claim 7, characterized in that: The cruise planning module includes an environment collection unit, a path planning unit, a dynamic optimization unit, and a backtracking adjustment unit. The environment collection unit is used to use sensors, sonar, high-definition cameras and lidar to collect real-time environmental data of the operating waters, and at the same time build a dynamic three-dimensional environmental model and update it in real time. The path planning unit is used to obtain the position information of the unmanned ship through Beidou or GPS navigation, and use the deep Q network reinforcement learning algorithm to perform initial path planning in combination with the environmental data. The position, speed and environmental information of the unmanned ship are used as state inputs to define discrete actions, design reward functions, guide the algorithm to learn path strategies, and consider multi-objective optimization to build a multi-objective function. The non-dominated sorting genetic algorithm is used to generate an initial path population. The dynamic optimization unit is used to navigate according to the initial path population. At set time intervals or when environmental data changes exceed a threshold, data is re-collected and the environmental model is updated. The deep Q network reinforcement learning algorithm is used again in combination with the real-time environmental status to adjust the path. At the same time, the non-dominated sorting genetic algorithm is used to dynamically optimize multiple objectives and select a navigation path. The backtracking adjustment unit is used to automatically backtrack after returning to normal when the planned path deviates due to a fault, compare the current position with the original planned path, and re-plan the path to the unfinished task area by combining real-time environmental data, reinforcement learning, and multi-objective optimization algorithms.

9. The intelligent cruising and collaborative operation system for a multifunctional unmanned vessel according to claim 7, characterized in that: The dynamic collaboration module includes a task collection unit, a priority determination unit, an evaluation and allocation unit, and a collaborative adjustment unit. The task collection unit is used to perform operations in the task area through various functional modules, and collect task data and external data at the same time. The priority determination unit is used to use a trained machine learning model to predict task priority, and at the same time, determine task priority in combination with preset rules. The evaluation and allocation unit is used to allocate resources based on task priority and the status of the unmanned ship using a multi-task scheduling algorithm, and execute tasks according to priority or in parallel. The collaborative adjustment unit is used to dynamically adjust the execution order and collaborative mode according to progress and resource occupancy during execution, and use sensors to monitor the status of each functional module, predict potential faults, and execute according to the emergency response plan.

Images