Unmanned ship control method and device, electronic equipment, storage medium and program product

Abstract

The invention provides an unmanned ship control method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of unmanned ship control, and the method comprises the steps: obtaining the operation data and environment data of an unmanned ship in a preset time period; detecting the system state of the unmanned ship based on the operation data to obtain a system state self-inspection result, obtaining a system fault prediction result based on the operation data and the environment data, and obtaining an environment risk assessment result based on the environment data; fusing the system state self-checking result and the system fault prediction result to obtain a system state fusion result; determining a starting strategy of the unmanned ship according to the environmental risk assessment result and the system state fusion result; and controlling each subsystem of the unmanned ship to execute corresponding operation according to the starting strategy. According to the invention, the misjudgment rate of remote starting of the unmanned ship can be reduced, and meanwhile, the starting safety and the starting success rate in a complex environment are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned vessel control technology, and in particular to an unmanned vessel control method, device, electronic equipment, storage medium, and program product. Background Technology

[0002] Unmanned surface vessels (USVs) are intelligent surface vessels that perform diverse tasks through autonomous navigation or remote control, primarily used in environmental monitoring, scientific research and exploration, civilian services, and military security.

[0003] Currently, unmanned surface vessels (USVs) are typically started remotely with a single click. A typical process involves the shore-based control unit establishing a remote communication connection with the USV and sending a one-click start command. The USV then executes a pre-set automated self-check program, determining whether start-up is permitted based on preset rules. However, the self-check method relies heavily on pre-set static thresholds to evaluate the USV's operational data. This approach struggles to effectively identify complex, potential, or sporadic faults, leading to risks of misjudgment and missed detections. Furthermore, the evaluation criteria are relatively simplistic, resulting in a higher risk of start-up failures and safety incidents. Summary of the Invention

[0004] This invention provides an unmanned vessel control method, device, electronic equipment, storage medium, and program product to address the shortcomings of poor reliability and safety in the startup control of unmanned vessels in the prior art.

[0005] This invention provides a method for controlling an unmanned surface vessel, comprising: Acquire operational and environmental data of the unmanned vessel within a preset time period; The system status of the unmanned vessel is detected based on the operational data to obtain the system status self-check result. Based on the operational data and the environmental data, the system fault prediction result is obtained, and based on the environmental data, the environmental risk assessment result is obtained. The system state self-check results are fused with the system fault prediction results to obtain the system state fusion result; Based on the environmental risk assessment results and the system state fusion results, the launch strategy for the unmanned vessel is determined; According to the activation strategy, control each subsystem of the unmanned vessel to perform corresponding operations.

[0006] According to an unmanned vessel control method provided by the present invention, the step of obtaining a system fault prediction result based on the operational data and the environmental data, and obtaining an environmental risk assessment result based on the environmental data, includes: The running data is converted into a running data sequence, and the environmental data is converted into an environmental data sequence; The system potential faults of the unmanned vessel are predicted using the fault prediction model based on the operational data sequence and the environmental data sequence, and the system fault prediction results are obtained. The environmental data is subjected to feature extraction and feature fusion processing to generate a fused feature map; The environmental risk is assessed using the fusion feature map based on the environmental risk assessment model, and the environmental risk assessment results are obtained.

[0007] According to the unmanned vessel control method provided by the present invention, the system fault prediction result includes the fault prediction probability of each subsystem; The process of fusing the system state self-check results with the system fault prediction results to obtain a system state fusion result includes: The self-test score of each subsystem is obtained based on the self-test results of the system status, and the model score of each subsystem is obtained based on the fault prediction probability of each subsystem. The self-inspection scores and model scores of each subsystem are weighted and summed to obtain the comprehensive health status score of each subsystem. The system status fusion result includes the comprehensive health status scores of each subsystem.

[0008] According to the unmanned vessel control method provided by the present invention, the environmental risk assessment result includes an environmental risk level; The step of determining the unmanned vessel launch strategy based on the environmental risk assessment results and the system state fusion results includes: Based on the system state fusion result, the system feature vector is obtained; The environmental risk level is numerically mapped to obtain a risk feature vector; The task feature vector, the system feature vector, and the risk feature vector are concatenated to obtain a state space vector; wherein, the task feature vector is obtained by encoding the task type. The launch strategy of the unmanned vessel is determined by using the launch decision model based on the state space vector.

[0009] According to the present invention, an unmanned surface vessel (USV) control method is provided, wherein controlling each subsystem of the USV to perform corresponding operations based on the activation strategy includes: If the startup strategy is full startup, then all subsystems controlling the unmanned vessel will be started. If the startup strategy is a function-restricted startup, then all subsystems controlling the unmanned vessel will be started, and the operating restrictions of the subsystems will be set. If the startup strategy is an incremental startup of core functions, then the core subsystems of the unmanned vessel are started, and after a preset time, the non-core subsystems are started. If the startup strategy is delayed startup, then the delay startup duration is obtained, and all subsystems of the unmanned vessel are started after the delay startup duration has elapsed.

[0010] According to the unmanned vessel control method provided by the present invention, before acquiring the unmanned vessel's operational data and environmental data within a preset time period, the method further includes: Establish a remote communication connection with the control terminal; A startup command is obtained based on the remote communication connection, and the startup command includes a digital signature; Verify the digital signature. If the verification is successful, perform the following steps: obtain the unmanned vessel's operational data and environmental data within a preset time period.

[0011] The present invention also provides an unmanned vessel control device, comprising the following modules: The data acquisition module is used to acquire the operational and environmental data of the unmanned vessel within a preset time period. The data processing module is used to detect the system status of the unmanned vessel based on the operational data, obtain the system status self-check result, obtain the system fault prediction result based on the operational data and the environmental data, and obtain the environmental risk assessment result based on the environmental data. The result fusion module is used to fuse the system state self-inspection results with the system fault prediction results to obtain the system state fusion result; The strategy determination module is used to determine the launch strategy of the unmanned vessel based on the environmental risk assessment results and the system state fusion results. The control module is used to control the various subsystems of the unmanned vessel to perform corresponding operations according to the launch strategy.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the unmanned vessel control methods described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the unmanned vessel control method as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the unmanned vessel control methods described above.

[0015] The unmanned surface vessel (USV) control method, device, electronic equipment, storage medium, and program products provided by this invention acquire operational and environmental data of the USV within a preset time period. While obtaining system status self-check results, the invention performs predictive diagnosis of potential subsystem faults and environmental risk assessment based on the operational and environmental data. Furthermore, it integrates the system status self-check results and system fault prediction results to comprehensively assess the health status of the subsystem. Compared to traditional self-checks based solely on operational data in existing technologies, this invention significantly improves the depth and breadth of self-checks, achieving a more accurate assessment of the subsystem's health status and reducing the occurrence of misjudgments and omissions. Simultaneously, by integrating environmental data into the system detection and startup decision-making process, this invention effectively avoids starting the USV in harsh or unsuitable environments, improving the safety of USV startup and increasing the startup success rate in complex environments (e.g., sea state 3). Moreover, the intelligent assessment and decision-making process of this invention significantly reduces the need for human intervention, lowering the risk of startup failure or safety accidents caused by human error or misjudgment. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the unmanned vessel control method provided by the present invention; Figure 2 This is a schematic diagram of the unmanned vessel control device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] Unmanned surface vessels (USVs) are intelligent surface vessels that perform diverse tasks through autonomous navigation or remote control, primarily used in environmental monitoring, scientific research and exploration, civilian services, and military security.

[0020] Currently, unmanned surface vessels (USVs) are typically started remotely with a single button. A typical process involves the shore-based control unit establishing a remote communication connection with the USV and sending a one-button start command; the USV then executes a pre-set automated self-check procedure and determines whether start-up is permitted based on pre-set rules.

[0021] However, existing technologies have limitations in the following aspects: Insufficient automation and intelligence in self-inspection: When unmanned vessels perform self-inspection using preset automated self-inspection procedures, their self-inspection methods mainly rely on using preset static thresholds to judge the unmanned vessel's operating data. This makes it difficult to effectively identify complex, potential, or sporadic faults, and there is a risk of misjudgment and missed judgment.

[0022] Lack of environmental adaptability: Existing startup methods mainly rely on operational data for judgment, which is a relatively singular dimension and generally lacks consideration of the environmental factors in which the unmanned vessel operates. This may lead to startup in complex or harsh environments, posing potential safety hazards. For example, forcibly starting the vessel when the wave height exceeds a preset threshold may damage the equipment.

[0023] Based on the above problems, this invention proposes an unmanned vessel control method, device, electronic equipment, storage medium, and program product, which are described below in conjunction with... Figures 1-3 Describe it.

[0024] Figure 1 This is one of the flowcharts illustrating the unmanned vessel control method provided by the present invention, such as... Figure 1 As shown, the method includes: Step S110: Obtain the operation data and environmental data of the unmanned vessel within a preset time period.

[0025] In this embodiment, the method can be applied to various unmanned vessel control scenarios.

[0026] In one type of unmanned vessel control scenario, the unmanned vessel control system includes an unmanned vessel and a control terminal, which communicate with each other via a remote communication link.

[0027] The control terminal, which can be a shore-based control terminal, is the core platform for operators to remotely control and monitor the status of the unmanned surface vessel (USV). Its hardware components include a high-performance workstation, monitor, human-machine interface devices (such as a keyboard, mouse, touchscreen, and biometric identification devices), and a network communication module. Its software system includes an operating system, remote control software, an authentication module, a data processing module, and a security encryption module. The remote control software, developed based on Qt (a cross-platform C++ application development framework) or Electron (a framework for building desktop applications using JavaScript, HTML, and CSS), provides a graphical user interface and enables secure communication with the USV, command transmission, status display, and data reception. The authentication module implements multi-factor authentication logic. The data processing module encapsulates the operator's activation commands into data packets conforming to the communication protocol and parses and displays the received USV data. The security encryption module uses TLS (Transport Layer Security) 1.3 protocol and AES-256 (an advanced encryption standard using a 256-bit key length) encryption algorithm to ensure communication security.

[0028] The remote communication link is responsible for data transmission between the shore-based control terminal and the unmanned vessel. It can employ various communication methods, such as mobile communication networks (4G / 5G), satellite communication, and wireless ad hoc networks. In offshore areas or areas without mobile network coverage, satellite communication can be used, specifically Iridium or maritime satellite communication systems (e.g., Cobham SAILOR Fleet Broadband). In near-shore or specific areas, wireless ad hoc networks can be used, specifically deploying ad hoc network communication schemes based on IEEE 802.11ah (also known as Wi-Fi HaLow, a low-power, long-range Wi-Fi standard) or LoRa (a low-power local area network wireless standard). The communication protocol stack uses TCP / IP (Transmission Control Protocol / Internet Protocol). The application layer protocol can be custom or use standard protocols, such as MQTT (Message Queuing Telemetry Transport) and WebSockets (a full-duplex communication protocol based on TCP). To ensure security, the communication link layer uses the TLS 1.3 protocol and is configured with a high-strength encryption suite, such as TLS_AES_256_GCM_SHA384.

[0029] The hardware components of an unmanned surface vessel (USV) include, but are not limited to, a main control unit, environmental sensors, and subsystems. The software system may include, but is not limited to, a data acquisition module, a data processing module, a result fusion module, a strategy determination module, and a control module. The main control unit can be an industrial-grade embedded computer running a real-time operating system. Environmental sensors include, but are not limited to, lidar, cameras, weather stations, and water quality sensors. The purpose of each module can be referred to in the embodiments of the electronic equipment described below.

[0030] In this application scenario, the unmanned surface vessel (USV) executes the scheme described in this embodiment when it receives a start command from the control terminal. That is, in this application scenario, the executing entity is the USV.

[0031] In another unmanned surface vessel (USV) control scenario, the USV control system includes the USV, a control terminal, and a cloud platform. The control terminal and the cloud platform communicate via a remote communication link, and the cloud platform and the USV communicate via the same link. In this scenario, the control terminal is only responsible for operator interaction and sending one-click start commands. Upon receiving the start command, the cloud platform executes the scheme described in this embodiment. The USV receives control commands from the cloud platform and controls its various subsystems to perform corresponding operations based on these commands. In other words, in this scenario, the cloud platform is the primary execution entity. This implementation method improves the system's scalability and computing power, and facilitates model updates and maintenance.

[0032] In this embodiment of the invention, the first application scenario described above is used as an example, in which the executing entity is an unmanned vessel.

[0033] Specifically, firstly, the operational and environmental data of the unmanned surface vessel (USV) within a preset time period are acquired. This data acquisition can occur upon receiving a start command, which can originate from the control unit or be a pre-set command within the USV that is triggered according to rules (such as a timer).

[0034] The preset time period can be set in advance as needed, for example, it can be set to a time period of 10 minutes before the current time, 20 minutes before the current time, without specific limitations. The unmanned vessel can periodically acquire operational and environmental data and store them in the database, for example, every 30 seconds or 1 minute. Then, when data within the preset time period is needed, it can be retrieved directly from the database.

[0035] The operational data of an unmanned surface vessel (USV) originates from its subsystems, which may include, but are not limited to, the following: a propulsion system, a navigation and positioning system, a remote communication system, a power management system, and a mission payload system. The propulsion system includes a brushless DC motor, an electronic speed controller, a propeller, and servo motors. The navigation and positioning system integrates a GNSS (Global Navigation Satellite System) receiver, an IMU (Inertial Measurement Unit), and a magnetometer, and employs a Kalman filter algorithm to fuse multi-sensor data, improving positioning accuracy and robustness. The remote communication system is the module that communicates with the shore-based control unit, using the same communication method. The power management system includes a lithium-ion battery pack, a battery management system, and a DC-DC (Direct Current-Direct Current) power converter, which monitors battery status in real time and distributes power accordingly. The mission payload system is used to configure appropriate payloads according to specific mission requirements, such as payloads for environmental monitoring and scientific research, surveying and exploration, and material transportation and operations.

[0036] Correspondingly, the operational data may include, but is not limited to, the following: motor current, voltage, speed, and temperature detected by the propulsion system, as well as propeller speed sensor data; the number of satellites, signal-to-noise ratio, and positioning accuracy of the GNSS receiver detected by the navigation and positioning system, as well as the three-axis angular velocity and acceleration detected by the IMU; the signal strength, link delay, and packet loss rate of the wireless communication module detected by the remote communication system; the battery voltage, current, and temperature detected by the power management system, as well as the operating status of each DC-DC converter; and the corresponding parameter detection performed by the mission payload system according to the specific payload configuration. These parameters are obtained around the following core objectives: ① Functional integrity check: confirming that the payload equipment itself and its components can work normally; ② Data accuracy prediction: checking whether the sensors need calibration or whether there are obvious abnormal readings; ③ Interface and communication status: ensuring smooth connection and data exchange between the payload and the unmanned vessel's main control system and power system; ④ Resource and environmental conditions: such as checking whether the storage space and power supply meet the requirements. For example, when it comes to material transport and operational payloads, the operational data of the mission payload system may include the cargo compartment door closure status, cargo temperature and humidity, sampling the status of motors or hydraulic systems of each joint of the robotic arm, limit switch status, gripper or tool head status, and the umbilical cable power supply and communication status of the underwater robot.

[0037] Environmental data includes one or more of meteorological data, water quality data, and surface object data, which can be collected using various environmental sensors mounted on unmanned surface vessels (USVs). Meteorological data includes, but is not limited to, wind speed, wind direction, temperature, humidity, atmospheric pressure, wave height, and visibility, and can be collected using weather stations (e.g., Vaisala WXT520). Water quality data includes, but is not limited to, pH value, dissolved oxygen, conductivity, and turbidity, and can be collected using water quality sensors (e.g., YSI EXO2). Surface object data includes, but is not limited to, 3D point cloud data and surface object images. Surface objects refer to objects floating or anchored on the water surface, including, but not limited to, static and dynamic obstacles, navigation marks, other vessels, and shorelines. Navigation marks are markers indicating the direction, boundaries, and obstructions of navigation channels, including, but not limited to, lighthouses, beacons, lighthouses, buoys, lightships, and various guide beacons. Specifically, 3D point cloud data can be obtained using lidar (e.g., Velodyne Puck LITE); surface object images can be obtained using industrial cameras (e.g., Basler ace series).

[0038] Step S120: Detect the system status of the unmanned vessel based on the operational data to obtain the system status self-check result; obtain the system fault prediction result based on the operational data and the environmental data; and obtain the environmental risk assessment result based on the environmental data.

[0039] The system performs a self-check on the operational data to obtain the system status self-check results. The specific self-check method can be to detect the operational data of each subsystem in parallel according to preset thresholds and rules. The operational data used for self-checking can be real-time operational data (i.e., the operational data at the last moment within a preset time period) or all operational data within a preset time period. Different self-check rules can be set for the operational data of different subsystems and different types of operational data used for self-checking. The self-check rules can be set according to the detection needs. For example, for the battery voltage of the propulsion system, if the self-check is for real-time operational data, the self-check rule can be set to determine an anomaly if the real-time battery voltage is lower than 12V; if the self-check is for operational data within a preset time period, the self-check rule can be set to determine an anomaly if the battery voltage is lower than 12V within the preset time period. The self-check rules here are only examples and are not specifically limited. By executing the self-check process of each subsystem in parallel, the self-check time can be shortened, thereby improving startup efficiency. In addition, by setting different self-check rules based on different operating data, the accuracy of the system status self-check results can be further improved, thereby improving the accuracy of subsequent startup strategies.

[0040] While performing self-checks, system fault prediction results are obtained based on operational and environmental data, and environmental risk assessment results are obtained based on environmental data. As one implementation method, a fault prediction model can be used to assess potential system faults of the unmanned vessel based on operational and environmental data to obtain system fault prediction results, and an environmental risk assessment model can be used to assess environmental risks based on environmental data to obtain environmental risk assessment results.

[0041] Step S130: The system state self-check result and the system fault prediction result are fused to obtain the system state fusion result.

[0042] The system state self-check results and system fault prediction results are fused to obtain the system state fusion result. The fusion methods may include, but are not limited to: 1) determining a self-check score based on the system state self-check results and a preset mapping table between self-check results and scores, determining a model score based on the system fault prediction results and a preset mapping table between fault prediction probabilities and scores, and summing the self-check score and the model score to obtain the system state fusion result; 2) determining a self-check score based on the system state self-check results and a preset mapping table between self-check results and scores, determining a model score based on the system fault prediction results and a preset mapping table between fault prediction probabilities and scores, and weighted summing the self-check score and the model score to obtain the system state fusion result.

[0043] Step S140: Determine the launch strategy for the unmanned vessel based on the environmental risk assessment results and the system state fusion results.

[0044] After obtaining the environmental risk assessment results and system state fusion results, the launch strategy for the unmanned vessel is determined. The methods for determining the launch strategy include, but are not limited to: 1) matching and determining the corresponding launch strategy based on the preset mapping relationship between different results and launch strategies; 2) determining the launch strategy based on the environmental risk assessment results, system state fusion results, and launch decision model.

[0045] Step S150: According to the startup strategy, control each subsystem of the unmanned vessel to perform corresponding operations.

[0046] After determining the launch strategy, the corresponding control operations are determined based on the launch strategy to control the execution of each subsystem of the unmanned vessel.

[0047] The unmanned surface vessel (USV) control method provided by this invention acquires operational and environmental data of the USV within a preset time period. While obtaining system status self-check results, it performs predictive diagnosis of potential subsystem faults and environmental risk assessment based on the operational and environmental data. Furthermore, it integrates the system status self-check results and system fault prediction results to comprehensively evaluate the health status of the subsystem. Compared to traditional self-checks based solely on operational data in existing technologies, this invention significantly improves the depth and breadth of self-checks, achieving a more accurate health status assessment of the subsystem and reducing the occurrence of misjudgments and omissions. Simultaneously, by integrating environmental data into the system detection and startup decision-making process, this invention effectively avoids starting the USV in harsh or unsuitable environments, improving the safety of USV startup and increasing the startup success rate in complex environments (e.g., sea state 3). Moreover, this invention significantly reduces the need for human intervention through intelligent evaluation and decision-making processes, lowering the risk of startup failure or safety accidents caused by human error or misjudgment.

[0048] In some implementations, the step "obtaining system fault prediction results based on the operational data and the environmental data, and obtaining environmental risk assessment results based on the environmental data" includes: Step S121: Convert the running data into a running data sequence and convert the environmental data into an environmental data sequence; Step S122: Using a fault prediction model, predict the potential system faults of the unmanned vessel based on the operational data sequence and the environmental data sequence, and obtain the system fault prediction results; Step S123: Perform feature extraction and feature fusion processing on the environmental data to generate a fused feature map; Step S124: Use the environmental risk assessment model to assess the environmental risk based on the fusion feature map to obtain the environmental risk assessment result.

[0049] It should be noted that the execution order of steps S121-S122 and steps S123-S124 is not important; that is, the steps of obtaining the system fault prediction results and the environmental risk assessment results can be executed in parallel to improve efficiency.

[0050] Specifically, the process of obtaining system fault prediction results is as follows: The running data is normalized, and a running data sequence is constructed based on the normalized running data. During normalization, each feature value can be scaled to the [0,1] interval. Furthermore, before normalization, the running data can be preprocessed, including noise removal and / or missing value imputation. For noise removal, outliers or noisy data can be smoothed (e.g., by moving average) or directly deleted. For missing values, interpolation methods (e.g., linear interpolation, polynomial interpolation) or the previous value can be used to fill in the missing data. Through the above preprocessing and normalization, data quality can be improved, which in turn helps improve the accuracy of the model output results.

[0051] While processing operational data, environmental data can be converted into an environmental data sequence. Specifically, the environmental data is first preprocessed, including but not limited to: 1) Time synchronization: using PTP (Precision Time Protocol) for nanosecond-level timestamp synchronization; 2) Coordinate system one: converting the environmental data to a geographic coordinate system based on WGS-84 (World Geodetic System-1984 Coordinate System) to align with the operational data of the subsystem; 3) Anomaly handling: using Kalman filtering to remove sensor noise and performing statistical outlier detection; 4) Normalization: scaling the feature values ​​of each environmental data point to the [0,1] range. Then, an environmental data sequence is constructed based on the preprocessed environmental data. The purpose of the above processing is to form an environmental data sequence with the same time step as the subsystem operational data sequence. It should be noted that, considering that meteorological data and water quality data often directly or indirectly affect the operating status and lifespan of each subsystem, and may even induce specific types of failures, environmental data sequences can be obtained by converting meteorological data and water quality data.

[0052] The operational data sequence and environmental data sequence are input into the fault prediction model to obtain the system fault prediction results. The fault prediction model is built based on a Long Short-Term Memory (LSTM) network. Compared to other deep learning models, choosing to build and train a fault prediction model using an LTM network allows for better processing of long-sequence data (including operational and environmental data sequences), enabling the extraction of effective features and thus resulting in more accurate system fault predictions. The system fault prediction results represent the probability distribution of specific faults (e.g., motor overheating, battery failure, GPS signal loss) occurring in each subsystem within the next 30 minutes, including the fault prediction type and probability for each subsystem.

[0053] The training process for the above fault prediction model is as follows: a1) Obtaining the training dataset: Collect operational data and fault records from multiple similar unmanned surface vessels (USVs) over the past five years, totaling approximately 10,000 hours of data, along with corresponding environmental data. The data originates from actual shipboard testing and simulation environments, covering normal operating conditions and various fault scenarios to ensure data diversity and comprehensiveness. Fault types include, but are not limited to: motor overheating, abnormal battery voltage, GPS signal loss, sensor failure, and communication interruption. Each fault type contains at least 500 instances to ensure the model can learn the characteristics of various faults.

[0054] a2) Data Preprocessing: For the original time series data, noise and missing values ​​are first removed, and then normalization is performed to scale each feature value to the [0,1] interval. Next, a sliding window method is used to generate training samples, with a window size of 10 minutes and a step size of 1 minute. The label of each sample is the fault type at the end of the window.

[0055] a3) Model Training: The TensorFlow 2.x framework (a deep learning framework) was used for model building and training. This framework provides rich application programming interfaces and efficient computing power, making it suitable for handling large-scale datasets. The model adopts a three-layer LSTM (Long Short-Term Memory) structure, with each layer containing 128 neurons. The Adam optimizer (an optimization algorithm based on gradient descent) was used, with an initial learning rate of 0.001 and a learning rate decay strategy (decreasing by 0.1 times every 10 epochs). The Adam optimizer combines the advantages of momentum and adaptive learning rates, enabling adaptive adjustment of the learning rate to improve model training efficiency. Through learning rate decay, rapid convergence can be achieved in the early stages of training, and model parameters can be finely adjusted in the later stages to avoid overfitting. The batch size was set to 64, which achieved a good balance between computational efficiency and model performance. The classification cross-entropy loss function was used. The training process lasted for 50 epochs (training rounds), and model performance was monitored using a validation set to prevent overfitting. The validation set accounted for 20% of the total dataset.

[0056] a4) Model Evaluation: Precision, recall, F1 score (balanced F-score), and AUC (Area Under Curve, the area under the receiver operating characteristic curve and the coordinate axis) were used as evaluation metrics for model performance. On the test set, the fault prediction model achieved an average prediction accuracy of over 90% for various common faults.

[0057] Specifically, the process of obtaining the results of the environmental risk assessment model is as follows: Feature extraction is performed on environmental data. Specifically, the 3D point cloud data from the water surface object data is voxelized and projected to convert it into a depth image; the water surface object images are cropped and scaled; simultaneously, meteorological and water quality data are normalized and then mapped to different pixel value ranges in the RGB (red-green-blue) channels. Then, a deep learning model based on an attention mechanism is used for feature fusion to generate a fused feature map containing comprehensive environmental state information. This fused feature map can be a two-dimensional feature map, such as 256x256 pixels. Through the deep learning model based on an attention mechanism, multi-source, heterogeneous environmental data can be intelligently fused and mapped to different channels or regions of the 256x256 feature map, optimizing the quality of the generated fused feature map and improving the accuracy of the environmental risk assessment model.

[0058] It should be noted that before feature extraction and feature fusion, environmental data can be preprocessed. Preprocessing methods include, but are not limited to: 1) Time synchronization: using PTP (Precision Time Protocol) for nanosecond-level timestamp synchronization; 2) Coordinate system transformation: converting environmental data to a geographic coordinate system based on WGS-84 (World Geodetic System-1984 Coordinate System) to align with the corresponding operational data of the subsystem; 3) Anomaly handling: using Kalman filtering to remove sensor noise and performing statistical outlier detection. These preprocessing steps improve data quality, thereby enhancing the accuracy of the model's output.

[0059] The fused feature map is input into the environmental risk assessment model, which outputs the environmental risk assessment result, including the environmental risk type and environmental risk level. The environmental risk assessment model is built based on a convolutional neural network. Compared to other deep learning models, choosing to build and train a fault prediction model using a convolutional neural network leverages its spatial feature extraction capabilities, parameter efficiency, and multimodal data fusion advantages, resulting in more accurate environmental risk assessment results.

[0060] The training process for the above environmental risk assessment model is as follows: b1) Construction of the training dataset: Environmental data collected from various environmental sensors mounted on the unmanned surface vessel (USV) under different sea states and weather conditions were gathered. Data sources included actual shipboard sea test records, open data from meteorological agencies, and hydrological data. Sea state levels were divided into 1-6, and weather conditions included sunny, cloudy, overcast, light rain, heavy rain, and fog. Each environmental state contained at least 300 instances. Environmental data included meteorological data, water quality data, and surface object data. Meteorological data included wind speed, wind direction, temperature, humidity, atmospheric pressure, wave height, and visibility; water quality data included pH, dissolved oxygen, conductivity, and turbidity; and surface object data included 3D point cloud data and surface object images. The acquired data was labeled by experienced navigators, categorized into low, medium, and high risk levels, and potential risk types (e.g., collision risk, grounding risk, severe weather risk) were also labeled.

[0061] b2) Data Processing: Environmental data obtained from multiple environmental sensors undergoes feature extraction and fusion to generate a 256x256 feature map, which serves as the model input. Specific steps are as follows: 3D point cloud data is voxelized and projected to convert it into a depth image. Water surface object images are cropped and scaled; unnecessary parts are removed by cropping, and the image size is adjusted to fit the model's input requirements. Meteorological and water quality data are normalized and mapped to different pixel value ranges in the RGB channels. Then, a deep learning model based on an attention mechanism is used for feature fusion to generate a 256x256 fused feature map. It should be noted that environmental data can be preprocessed before feature extraction and fusion. Preprocessing methods include, but are not limited to: 1) time synchronization; 2) coordinate system unification; 3) anomaly handling. Specific preprocessing procedures can be found in the above embodiments and will not be elaborated here.

[0062] b3) Model Training: The PyTorch framework (a deep learning framework) was used for model building and training. PyTorch is suitable for handling complex deep learning tasks. The model adopts a five-layer Convolutional Neural Network (CNN) structure with a kernel size of 3x3, a stride of 1, a ReLU activation function, and max pooling. The Adam optimizer was used with an initial learning rate of 0.0005, and an early stopping strategy was employed. Training was stopped when the validation set loss did not decrease for 10 consecutive epochs. This strategy effectively prevents overfitting and ensures the model's performance on the validation set. The batch size was set to 32. The classification cross-entropy loss function was used. The training process lasted for 100 epochs, iterating and optimizing the model parameters to ensure that the model could learn the features in the data.

[0063] b4) Model Evaluation: Precision, recall, F1 score, and confusion matrix were used as evaluation metrics for model performance. On the test set, the environmental risk assessment model achieved a classification accuracy of over 85% for environmental risk levels.

[0064] In this embodiment, the fault prediction model built on LSTM can better process long-sequence data (including operational data sequences and environmental data sequences), filter and extract effective features, and achieve more accurate predictive diagnosis of potential system faults of the unmanned vessel. This model can then be used in conjunction with the system status self-inspection results to comprehensively assess the health status of the subsystem, enabling a more accurate health status assessment of the unmanned vessel and further reducing the occurrence of false positives and false negatives. Simultaneously, by utilizing an environmental risk assessment model built on CNN, its spatial feature extraction capabilities, parameter efficiency, and multimodal data fusion advantages can be leveraged to make the environmental risk assessment results more accurate. This effectively avoids starting the unmanned vessel in harsh or unsuitable environments, further improving the success rate of startup in complex environments (such as sea state 3).

[0065] In some implementations, the system fault prediction result includes the fault prediction probability of each subsystem, and step S130 includes: Step S131: Obtain the self-test score of each subsystem based on the system status self-test results, and obtain the model score of each subsystem based on the fault prediction probability of each subsystem. Step S132: The self-inspection score and the model score of each subsystem are weighted and summed to obtain the comprehensive health status score of each subsystem. The system status fusion result includes the comprehensive health status score of each subsystem.

[0066] Specifically, in step S130, the system state self-check result and the system fault prediction result are fused to obtain the system state fusion result, which may include: The self-inspection score of each subsystem is obtained based on the system status self-inspection results and the preset mapping table between self-inspection results and scores. The preset mapping table between self-inspection results and scores can be specifically set according to the type of subsystem, the type of subsystem's operating data, and the evaluation method of its self-inspection results, without specific limitations.

[0067] Simultaneously, model scores for each subsystem are obtained based on the fault prediction probability. These model scores represent the health status assessment results obtained based on artificial intelligence. The model scores can be determined based on the fault prediction probability and a preset mapping table between the fault prediction probability and the score, or they can be calculated according to a preset formula. For example, the preset formula can be set as: Model score = 1 - Fault prediction probability.

[0068] Then, the self-inspection scores and model scores of each subsystem are weighted and summed to obtain the comprehensive health status score of each subsystem.

[0069] Based on the above, the specific calculation formula can be: Among them, S health It is a comprehensive health status score, W traditional and W model It is a preset weighting coefficient, S traditional It is a self-assessment score, S model It is the model score, P fault It is the probability of fault prediction.

[0070] It should be noted that, for W traditional and W model These two preset weighting coefficients can be set differently for different subsystems. Specifically, the weighting coefficients can be set based on the importance of the subsystem and the confidence level of the fault prediction model. The aim is to maximize the safety of the unmanned vessel and ensure that the health status of core subsystems affecting navigation safety receives the most rigorous assessment and consideration. For example, core subsystems (such as the propulsion system and navigation and positioning system) can be assigned a higher weighting coefficient than non-core subsystems. This gives the comprehensive health status score of the core subsystem a higher decision-making weight or influence, meaning it has a higher weight in subsequent startup decisions, thereby improving the accuracy of the startup strategy. As another example, for any subsystem, if the fault prediction model has been verified to be highly reliable and can provide early warnings of critical faults, then W can be set... traditional <W model This ensures that the model's predictions dominate its comprehensive health status assessment; if traditional self-assessment results are considered an absolutely reliable baseline, then W can be set. traditional >W model This ensures the decisive role of traditional testing. By employing the above methods, the sensitivity of the comprehensive health status assessment to potential risks can be ensured, and the reliance on traditional self-inspection methods and model predictions can be balanced. This maximizes the safety of unmanned vessels and ensures that the health status of core subsystems affecting navigation safety receives the most rigorous assessment and consideration.

[0071] In this embodiment, the system status self-inspection results obtained by each subsystem based on traditional self-inspection methods are weighted and fused with the model evaluation results (system fault prediction results) obtained based on artificial intelligence to form a system status fusion result that includes the comprehensive health status score of each subsystem. Compared with the prior art that only relies on traditional system status self-inspection results, the present invention can more accurately assess the health status of each subsystem, significantly improve the depth and breadth of traditional self-inspection, and reduce the occurrence of misjudgment and omission.

[0072] In some embodiments, the environmental risk assessment result includes an environmental risk level, and step S140 includes: Step S141: Obtain the system feature vector based on the system state fusion result; Step S142: Perform numerical mapping on the environmental risk level to obtain a risk feature vector; Step S143: Concatenate the task feature vector, the system feature vector, and the risk feature vector to obtain a state space vector; wherein, the task feature vector is obtained by encoding the task type. Step S144: Using the startup decision model, determine the startup strategy of the unmanned vessel based on the state space vector.

[0073] Given that existing technologies typically employ an "either / or" startup strategy—meaning startup is prohibited if any critical self-check item fails to meet preset conditions—they lack the ability to handle situations with finer precision based on fault severity and environmental risk. Even minor malfunctions in non-critical sensors can halt the entire startup process. Therefore, this invention generates multiple flexible startup strategies based on a startup decision model to maximize mission completion while ensuring the safe operation of the unmanned vessel, thereby improving the flexibility and efficiency of mission execution.

[0074] It should be noted that the execution order of steps S141 and S142 is not important and they can be executed in parallel.

[0075] Specifically, based on the system state fusion results, a system feature vector is obtained. This fusion result includes the comprehensive health status score of each subsystem. The health status scores of the subsystems can be normalized, scaling the data to the range [0,1] to obtain the system feature vector. Simultaneously, environmental risk levels are numerically mapped; for example, a low environmental risk level is mapped to 0.25, a medium environmental risk level to 0.5, and a high environmental risk level to 0.75, to obtain a risk feature vector. Next, task feature vectors are obtained. Specifically, task types, such as patrol, surveying, and transportation, are first identified, and then encoded using One-Hot encoding to obtain task feature vectors. Then, the task feature vector, system feature vector, and risk feature vector are concatenated to obtain a state space vector. This state space vector is then input into a pre-trained startup decision model to obtain the unmanned vessel's startup strategy. The startup decision model is constructed based on proximal policy optimization algorithms (PPO).

[0076] In this embodiment, by using the subsystem's comprehensive health status score, environmental risk level, and task type as the state space input of the startup decision model, the factors affecting the startup of the unmanned vessel are fully considered, making the generated startup decision more targeted and adaptable, and better serving diverse system health status situations, environmental risk situations, and task scenarios.

[0077] Specifically, the training process for this startup decision model is as follows: c1) Construction of state space and action space: The state space includes: the comprehensive health status score of each subsystem (between 0 and 1), the environmental risk level (including low, medium and high, which can be mapped to 0.25, 0.5 and 0.75 respectively), and the task type (including patrol, survey, transportation, etc., represented by One-Hot encoding).

[0078] The action space includes five activation strategies: ① full activation, ② function-restricted activation, ③ core function incremental activation (prioritizing the activation of navigation and propulsion systems), ④ delayed activation, and ⑤ activation prohibited.

[0079] c2) Reward Function Design: The design of the reward function is crucial, as it directly affects the learning direction and final performance of the agent. The reward function should comprehensively consider the following factors: ① Mission completion reward: Successfully reaching the target area will grant a positive reward (e.g., +10), while failing to reach it will result in a penalty (e.g., -1).

[0080] ② Time penalty: A slight negative reward (e.g., -0.001) is given for each time step to encourage the agent to complete the task as soon as possible.

[0081] ③ Energy consumption penalty: A negative reward (e.g., -0.01) is given for each unit of electricity consumed to encourage agents to choose energy-saving startup strategies.

[0082] ④ Safety status reward / penalty: A large negative penalty (e.g., -100) is given for a collision or serious failure, and a slight positive reward (e.g., +0.001) is given for continuous safe operation every minute.

[0083] ⑤ Startup efficiency bonus: A slight positive bonus (e.g., +0.1) is given for choosing a more efficient startup strategy (e.g., incremental startup of core functions).

[0084] c3) Model Training: The PPO algorithm was implemented using the Stable Baselines3 library. The model employs an Actor-Critic structure, with both the Actor (policy network) and Critic (value network) using a three-layer fully connected neural network, each layer containing 64 neurons and using the ReLU activation function. The learning rate was 0.0003. The discount factor (γ) was set to 0.99. Gae_lambda was set to 0.95. The training process lasted for 1 million time steps.

[0085] c4) Model Evaluation: The performance of the reinforcement learning model is evaluated through multiple tests in a simulation environment. Evaluation metrics include average reward, task completion rate, and average startup time.

[0086] It should be noted that, based on the comprehensive health status score and environmental risk level of the subsystems, the task type is added as a state input. Furthermore, the reward function design considers task-related requirements such as task area, task objective, time constraints, and energy budget. This allows the startup decision model to learn during training which subsystem health status combinations and environmental risk levels should be adopted for different task types. After this training, the startup strategy model can select appropriate startup strategies based on different system state fusion results, environmental risk levels, and task types, and complete tasks as efficiently as possible while ensuring the safety of the unmanned surface vessel.

[0087] In some implementations, step S150 includes: Step S151: If the startup strategy is full startup, then all subsystems controlling the unmanned vessel will be started. Step S152: If the startup strategy is a function-restricted startup, then all subsystems controlling the unmanned vessel are started, and the operating restrictions of the subsystems are set. Step S153: If the startup strategy is incremental startup of core functions, then control the startup of the core subsystem of the unmanned vessel, and after a preset time, control the startup of the non-core subsystem. Step S154: If the startup strategy is delayed startup, then obtain the delayed startup duration, and control all subsystems of the unmanned vessel to start up after the delayed startup duration has elapsed.

[0088] Specifically, startup strategies may include, but are not limited to: full startup, feature-restricted startup, incremental startup of core functions, delayed startup, and startup disabled.

[0089] If the startup strategy is full startup, all subsystems controlling the unmanned vessel will be started. All functional modules carried by all subsystems will operate with their designed full performance indicators without any restrictions, without actively restricting or degrading any function.

[0090] Understandably, to save energy and improve startup efficiency, the task type can be obtained first, the target subsystem can be determined based on the task type, and all target subsystems can be started. That is, only the target subsystems required to execute the task can be started.

[0091] If the startup strategy is a function-restricted startup, this strategy typically allows all subsystems to start, even if a non-core subsystem has a minor fault (e.g., an auxiliary sensor reading is unstable but does not affect core safety). This allows the non-core subsystem to start, making full use of its remaining functionality or providing auxiliary information to other subsystems, while restricting the use of some non-critical functions. This startup strategy, while ensuring the core mission and the safety of the unmanned surface vessel (USV), allows the USV to be put into operation as much as possible, even with some controllable secondary problems or risks. However, it imposes corresponding constraints on affected non-critical functions or overall performance indicators, thereby improving the flexibility and efficiency of USV mission execution.

[0092] Specifically, all subsystems controlling the unmanned vessel are activated, and operational constraints are set for each subsystem according to preset restriction rules. These preset restriction rules include various pre-defined faults for each subsystem and their corresponding restrictions. In setting operational constraints, in addition to the preset restriction rules, system self-check results and / or system fault prediction results and / or environmental risk assessment results can also be considered. For example, if a backup link in the remote communication system has a weak signal (damaging some functions of non-critical subsystems), operational constraints can be set to limit the data types transmitted through that link or reduce the clarity of video transmissions to lower its performance indicators. Similarly, if the propulsion system itself is healthy, but the environmental risk assessment result is "moderate to controllable" (e.g., wave height slightly higher than ideal), operational constraints can be set to limit the unmanned vessel's maximum speed or acceleration to lower the performance indicators of the propulsion system.

[0093] Similarly, to save energy and improve startup efficiency, not all subsystems may be started, but only the target subsystems determined according to the task type may be started.

[0094] If the startup strategy is incremental startup of core functions, then core subsystems will be started first. Core subsystems are those essential for performing the current task, such as navigation and positioning systems, propulsion systems, remote communication systems, and mission payload systems directly related to the current task. Other non-core subsystems (such as mission payload systems not directly related to the current task) will be started with a delay. One implementation method is to start non-core subsystems after a preset time; another is to start them after the core subsystems are detected to be operating stably. Furthermore, when starting non-core subsystems, they can be started gradually and as needed based on task progress and system resource availability, rather than all at once. This strategy allows the unmanned surface vessel to quickly acquire the basic capabilities to perform its mission.

[0095] If the startup strategy is delayed startup, this strategy is typically employed when the self-check results of the unmanned surface vessel's (USV) critical subsystems (such as propulsion and navigation systems) are poor, or when there is a significant potential risk of failure (i.e., system failure prediction results indicate a high probability of a serious failure in the near future), or when the environmental risk assessment results show an excessively high environmental risk level (e.g., severe sea conditions or an impending storm). In this case, the system chooses not to start immediately, but rather to delay startup until human intervention or when the risk decreases. One implementation method involves obtaining the delayed startup duration, and controlling all USV subsystems to start after the duration has elapsed. The delayed startup duration can be determined based on the environmental risk assessment results and / or system state fusion results, or it can be the delayed startup duration fed back by the shore-based control terminal based on this delayed startup strategy. Another implementation method involves, upon receiving a startup command from the control terminal, controlling each subsystem of the USV to perform corresponding operations according to the startup command.

[0096] If the startup strategy is to prohibit startup, this strategy is usually due to a serious malfunction in the critical subsystem of the unmanned vessel or an extremely high environmental risk that may endanger the safety of the unmanned vessel. In this case, startup is prohibited, meaning that the various subsystems of the unmanned vessel are not controlled to perform any operations.

[0097] Furthermore, if the startup strategy is delayed startup or startup is disabled, corresponding alarm information is sent to the control terminal. This alarm information includes, but is not limited to: fault type, faulty subsystem, fault details, environmental risk assessment results, and recommended handling measures. Sending alarm information to the control terminal allows it to promptly understand the fault situation and take appropriate action.

[0098] In this embodiment, based on multiple startup strategies, corresponding operations are controlled to be performed on the unmanned vessel, which can maximize task completion while ensuring safety and improving the flexibility and efficiency of task execution. In addition, some startup strategies, such as incremental startup of core functions, adopt a batch startup approach for subsystems, which enables the unmanned vessel to quickly acquire the basic ability to perform tasks, achieve rapid response, and effectively shorten startup time.

[0099] Based on the above embodiments, prior to step S110, the unmanned vessel control method further includes: Step A: Establish a remote communication connection with the control terminal; Step B: Obtain a startup command based on the remote communication connection, the startup command including a digital signature; Step C: Verify the digital signature. If the verification is successful, execute step S110: Obtain the operation data and environmental data of each subsystem of the unmanned vessel within a preset time period.

[0100] Considering the insufficient security design of existing methods in the startup verification stage, there is a risk that the unmanned vessel may be started illegally. Therefore, this embodiment will verify the identity of the operator before controlling the start of the unmanned vessel to prevent illegal startup and enhance the security of the unmanned vessel startup.

[0101] In this embodiment, a remote communication link is first established between the unmanned surface vessel (USV) and the control terminal. The control terminal can be a remote control terminal, such as a shore-based control terminal, to remotely control the USV. Remote communication connection methods include, but are not limited to: wireless ad hoc networks, mobile communication networks (4G / 5G), satellite communication, etc.

[0102] Furthermore, to ensure communication security, a secure and reliable remote communication link based on the TLS 1.3 protocol can be established between the control terminal and the unmanned vessel, for example, by using a suite based on AES-256 encryption.

[0103] After establishing a remote communication connection, the operator can initiate a one-click start operation through the human-machine interface on the shore-based control terminal. The data processing module on the shore-based control terminal will create a data packet containing the start command, and then use the security encryption module to digitally sign the data packet using RSA (Rivest-Shamir-Adleman, an asymmetric encryption algorithm) or ECDSA (Elliptic Curve Digital Signature Algorithm) algorithms and the operator's private key to obtain a data packet with a digital signature containing user identity information. This data packet is then securely sent to the unmanned vessel through the established TLS 1.3 encrypted communication link. Correspondingly, after receiving the start command, the unmanned vessel obtains the digital signature in the start command and verifies it to verify the operator's identity. The specific digital signature verification method can refer to existing technologies. Upon successful verification, step S110 is executed: obtaining the unmanned vessel's operational data and environmental data within a preset time period, and then executing subsequent steps. The specific execution process can refer to the above embodiment and will not be elaborated here.

[0104] In this embodiment, before executing a series of control operations, the received startup command is authenticated, which can effectively prevent illegal startup and enhance system security.

[0105] Furthermore, it should be noted that after establishing a remote communication connection, the shore-based control terminal can first verify the operator's identity. Only after successful verification can the operator be allowed to start the unmanned vessel. The specific identity verification process at the shore-based control terminal can be as follows: the shore-based control terminal prompts the operator to verify their identity, and the verification methods include, but are not limited to, username / password combination verification, digital certificate authentication, and biometric recognition. Specifically, the operator can enter their username and password at the shore-based control terminal and submit them via HTTPS protocol. The shore-based control terminal uses the bcrypt algorithm (a password hashing algorithm) to encrypt and verify the password. Biometric recognition is performed using biometric recognition devices integrated into the shore-based control terminal. Biometric features can include, but are not limited to, fingerprint features, facial features, and voiceprint features. Taking fingerprint features as an example, the operator's fingerprint data can be collected through a fingerprint recognition module (e.g., an FPC1020 sensor using the Minutiae feature matching algorithm) and compared with pre-stored fingerprint data for verification. A digital certificate is used to prove the validity of a digital certificate held by an operator that conforms to the X.509 standard. Digital certificate authentication can be performed based on the OpenSSL (Open Secure Sockets Layer) library. Specific verification methods can refer to existing technologies, which will not be elaborated here.

[0106] By implementing a multi-factor authentication process at the shore-based control terminal, and only allowing the sending of start commands to the unmanned vessel after successful authentication, illegal start-up can be effectively prevented, thus enhancing the security of unmanned vessel start-up.

[0107] Based on the above embodiments, after step S150, the unmanned vessel control method further includes: Regularly monitor the operational status of the unmanned vessel; When an abnormal operating status is detected, the unmanned vessel's propulsion system is stopped, it enters a safe drifting mode, and a distress signal is sent.

[0108] In this embodiment, the operating status of the unmanned vessel can be monitored periodically during or after startup. The monitoring method can be to read data from the motor controller and / or battery management system via the CAN (Controller Area Network) bus and check whether the read monitoring data meets the preset requirements.

[0109] When monitoring detects that the unmanned vessel does not meet the preset requirements, i.e. when the operating status is abnormal, an emergency stop operation can be initiated. Specifically, the unmanned vessel's power propulsion system is stopped, the autopilot is activated, it enters a safe drifting mode, and a distress message is sent via satellite communication.

[0110] In this embodiment, by monitoring the operating status of the unmanned vessel, an emergency stop operation is automatically initiated when an abnormal operating status is detected, which can ensure the safety of the unmanned vessel's operation.

[0111] Based on the above embodiments, after step S150, the unmanned vessel control method further includes: Upon receiving a remote emergency stop command, the system controls the propulsion system to stop and enters a safe drifting mode.

[0112] During the operation of the unmanned vessel, the control terminal can send a remote emergency abort command with the highest priority. When the unmanned vessel receives the remote emergency abort command, it controls the power propulsion system to stop, activates the autopilot, and enters a safe drifting mode.

[0113] This remote emergency shutdown mechanism enhances the system's ability to respond to emergencies and ensures the safety of unmanned vessels.

[0114] Furthermore, after remote fault diagnosis or on-site maintenance to troubleshoot the fault, the control terminal can initiate a one-click start command again. At this time, after receiving the start command, the unmanned vessel can verify the digital signature included in the start command, and when the verification is successful, execute step S110: obtain the operating data and environmental data of each subsystem of the unmanned vessel within a preset time period, and then proceed with the subsequent steps. The specific execution process can be referred to the above embodiment, and will not be elaborated here.

[0115] To verify the effectiveness of the method of the present invention, we conducted a large number of simulation experiments and actual shipboard sea tests.

[0116] The proposed method was compared with traditional fixed-threshold-based startup methods under different sea states and fault scenarios. Experimental results show that the proposed method can significantly reduce the false alarm rate of remote startup of unmanned vessels by 15%-25%, improve the startup success rate in complex environments (such as sea state 3) by 10%-15%, and shorten the average startup time by 20%-30%.

[0117] Table 1 Simulation Experiment Results

[0118] In actual sea trials, multiple remote one-click start tests were conducted under different weather conditions and sea states. The test results were basically consistent with the simulation results, demonstrating the effectiveness and reliability of the method of this invention in real-world environments.

[0119] Table 2 Results of actual shipboard sea trials

[0120] It should be noted that the traditional fixed threshold-based startup method allows startup only when the running data of all subsystems meet the preset threshold range; otherwise, startup is prohibited.

[0121] In Tables 1 and 2 above, the false positive rate is calculated as follows: False positive rate = (Total number of false positives / Total number of test scenarios) * 100%. The total number of false positives is calculated based on the following two scenarios: ① False rejection: The actual state of the unmanned vessel allows it to start safely and is capable of performing its intended mission (or at least its core mission), but the startup decision method determines that it should not start, thus missing the opportunity to perform the mission; ② False acceptance: The actual state of the unmanned vessel (e.g., there is a potential for a critical failure, or the environmental risk is too high for startup) should not have allowed startup, but the startup decision method determines that it can start. This may lead to a failure shortly after startup, mission failure, or even more serious safety problems.

[0122] In Tables 1 and 2 above, the success rate of starting up in sea state 3 refers to the success rate of starting up the unmanned vessel under sea state 3. "Sea state 3" refers to a standardized sea surface condition level. Sea state levels are typically classified based on wind speed and the resulting wave height (e.g., referring to the Beaufort scale or Douglas scale). Sea state 3 usually corresponds to Beaufort scale 4 (gentle wind), with wind speeds of approximately 5.5-7.9 m / s and wave heights of approximately 0.5-1.25 meters. In this sea state, small waves and whitecaps are frequent. For small unmanned vessels, this is a challenging but generally still operable environment; therefore, the success rate of starting up under this sea state is used for performance comparison. Success rate of starting up in sea state 3 = (Number of times successfully started up under sea state 3 / Total number of test scenarios under sea state 3) * 100%. "Successful startup" means that the method allows startup and the unmanned surface vessel (USV) can operate stably in sea state 3. The criteria for stable operation include: ① successful completion of the startup procedure; ② key subsystems being able to operate stably for a period of time under the set sea state 3 disturbance (e.g., no attitude loss due to sea state, no continuous exceedance of key parameters, etc.); ③ ability to execute a preset simple navigation task (e.g., maintaining course, navigating to a nearby designated point). At this point, it is considered a "successful startup" in sea state 3.

[0123] The unmanned vessel control device provided by the present invention is described below. The unmanned vessel control device described below can be referred to in correspondence with the unmanned vessel control method described above.

[0124] Figure 2 This is a schematic diagram of the unmanned vessel control device provided by the present invention, as shown below. Figure 2 As shown, the device includes a data acquisition module 210, a data processing module 220, a result fusion module 230, a strategy determination module 240, and a control module 250; wherein: The data acquisition module 210 is used to acquire the operational data and environmental data of the unmanned vessel within a preset time period; The data processing module 220 is used to detect the system status of the unmanned vessel based on the operating data, obtain the system status self-check result, obtain the system fault prediction result based on the operating data and the environmental data, and obtain the environmental risk assessment result based on the environmental data. The result fusion module 230 is used to fuse the system state self-check result with the system fault prediction result to obtain the system state fusion result; The strategy determination module 240 is used to determine the launch strategy of the unmanned vessel based on the environmental risk assessment results and the system state fusion results. The control module 250 is used to control each subsystem of the unmanned vessel to perform corresponding operations according to the launch strategy.

[0125] The unmanned surface vessel (USV) control device provided by this invention acquires operational and environmental data of the USV within a preset time period. While obtaining system status self-check results, it performs predictive diagnosis of potential subsystem faults and environmental risk assessment based on the operational and environmental data. Furthermore, it integrates the system status self-check results and system fault prediction results to comprehensively evaluate the health status of the subsystems. Compared to traditional self-checks based solely on operational data in existing technologies, this invention significantly improves the depth and breadth of self-checks, achieving a more accurate health status assessment of the subsystems and reducing the occurrence of misjudgments and omissions. Simultaneously, by integrating environmental data into the system detection and startup decision-making process, this invention effectively avoids starting the USV in harsh or unsuitable environments, improving the safety of USV startup and increasing the startup success rate in complex environments (e.g., sea state 3). Moreover, this invention significantly reduces the need for human intervention through intelligent evaluation and decision-making processes, lowering the risk of startup failures or safety accidents caused by human error or misjudgment.

[0126] It should be noted that the unmanned vessel control device provided in this embodiment of the invention can implement all the method steps implemented in the above-mentioned unmanned vessel control method embodiment and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0127] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an unmanned vessel control method, which includes: Acquire operational and environmental data of the unmanned vessel within a preset time period; The system status of the unmanned vessel is detected based on the operational data to obtain the system status self-check result. Based on the operational data and the environmental data, the system fault prediction result is obtained, and based on the environmental data, the environmental risk assessment result is obtained. The system state self-check results are fused with the system fault prediction results to obtain the system state fusion result; Based on the environmental risk assessment results and the system state fusion results, the launch strategy for the unmanned vessel is determined; According to the activation strategy, control each subsystem of the unmanned vessel to perform corresponding operations.

[0128] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0129] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the unmanned vessel control method provided by the above methods, the method comprising: Acquire operational and environmental data of the unmanned vessel within a preset time period; The system status of the unmanned vessel is detected based on the operational data to obtain the system status self-check result. Based on the operational data and the environmental data, the system fault prediction result is obtained, and based on the environmental data, the environmental risk assessment result is obtained. The system state self-check results are fused with the system fault prediction results to obtain the system state fusion result; Based on the environmental risk assessment results and the system state fusion results, the launch strategy for the unmanned vessel is determined; According to the activation strategy, control each subsystem of the unmanned vessel to perform corresponding operations.

[0130] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the unmanned vessel control method provided by the methods described above, the method comprising: Acquire operational and environmental data of the unmanned vessel within a preset time period; The system status of the unmanned vessel is detected based on the operational data to obtain the system status self-check result. Based on the operational data and the environmental data, the system fault prediction result is obtained, and based on the environmental data, the environmental risk assessment result is obtained. The system state self-check results are fused with the system fault prediction results to obtain the system state fusion result; Based on the environmental risk assessment results and the system state fusion results, the launch strategy for the unmanned vessel is determined; According to the activation strategy, control each subsystem of the unmanned vessel to perform corresponding operations.

[0131] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0132] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling an unmanned surface vessel, characterized in that, include: Acquire operational and environmental data of the unmanned vessel within a preset time period; The system status of the unmanned vessel is detected based on the operational data to obtain the system status self-check result. Based on the operational data and the environmental data, the system fault prediction result is obtained, and based on the environmental data, the environmental risk assessment result is obtained. The system state self-check results are fused with the system fault prediction results to obtain the system state fusion result; Based on the environmental risk assessment results and the system state fusion results, the launch strategy for the unmanned vessel is determined; According to the activation strategy, control each subsystem of the unmanned vessel to perform corresponding operations.

2. The unmanned vessel control method according to claim 1, characterized in that, The process of obtaining system fault prediction results based on the operational data and the environmental data, and obtaining environmental risk assessment results based on the environmental data, includes: The running data is converted into a running data sequence, and the environmental data is converted into an environmental data sequence; The system potential faults of the unmanned vessel are predicted using the fault prediction model based on the operational data sequence and the environmental data sequence, and the system fault prediction results are obtained. The environmental data is subjected to feature extraction and feature fusion processing to generate a fused feature map; The environmental risk is assessed using the fusion feature map based on the environmental risk assessment model, and the environmental risk assessment results are obtained.

3. The unmanned vessel control method according to claim 1, characterized in that, The system fault prediction results include the fault prediction probabilities of each subsystem. The process of fusing the system state self-check results with the system fault prediction results to obtain a system state fusion result includes: The self-test score of each subsystem is obtained based on the self-test results of the system status, and the model score of each subsystem is obtained based on the fault prediction probability of each subsystem. The self-inspection scores and model scores of each subsystem are weighted and summed to obtain the comprehensive health status score of each subsystem. The system status fusion result includes the comprehensive health status scores of each subsystem.

4. The unmanned vessel control method according to claim 1, characterized in that, The environmental risk assessment results include the environmental risk level; The step of determining the unmanned vessel launch strategy based on the environmental risk assessment results and the system state fusion results includes: Based on the system state fusion result, the system feature vector is obtained; The environmental risk level is numerically mapped to obtain a risk feature vector; The task feature vector, the system feature vector, and the risk feature vector are concatenated to obtain a state space vector; wherein, the task feature vector is obtained by encoding the task type. The launch strategy of the unmanned vessel is determined by using the launch decision model based on the state space vector.

5. The unmanned vessel control method according to any one of claims 1 to 4, characterized in that, The step of controlling each subsystem of the unmanned vessel to perform corresponding operations according to the activation strategy includes: If the startup strategy is full startup, then all subsystems controlling the unmanned vessel will be started. If the startup strategy is a function-restricted startup, then all subsystems controlling the unmanned vessel will be started, and the operating restrictions of the subsystems will be set. If the startup strategy is an incremental startup of core functions, then the core subsystems of the unmanned vessel are started, and after a preset time, the non-core subsystems are started. If the startup strategy is delayed startup, then the delay startup duration is obtained, and all subsystems of the unmanned vessel are started after the delay startup duration has elapsed.

6. The unmanned vessel control method according to any one of claims 1 to 4, characterized in that, Before acquiring the unmanned vessel's operational and environmental data within a preset time period, the process also includes: Establish a remote communication connection with the control terminal; A startup command is obtained based on the remote communication connection, and the startup command includes a digital signature; Verify the digital signature. If the verification is successful, perform the following steps: obtain the unmanned vessel's operational data and environmental data within a preset time period.

7. An unmanned surface vessel control device, characterized in that, include: The data acquisition module is used to acquire the operational and environmental data of the unmanned vessel within a preset time period. The data processing module is used to detect the system status of the unmanned vessel based on the operational data, obtain the system status self-check result, obtain the system fault prediction result based on the operational data and the environmental data, and obtain the environmental risk assessment result based on the environmental data. The result fusion module is used to fuse the system state self-inspection results with the system fault prediction results to obtain the system state fusion result; The strategy determination module is used to determine the launch strategy of the unmanned vessel based on the environmental risk assessment results and the system state fusion results. The control module is used to control the various subsystems of the unmanned vessel to perform corresponding operations according to the launch strategy.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the unmanned vessel control method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the unmanned vessel control method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the unmanned vessel control method as described in any one of claims 1 to 6.

Images