Control method of unmanned ship

Abstract

The invention relates to the technical field of unmanned ship control, in particular to an unmanned ship control method, which is used for upgrading a sensing device of an unmanned ship so as to obtain more accurate environmental data. According to the characteristics of unmanned ship control, a deep learning algorithm is designed and optimized, so that the deep learning algorithm can more effectively extract characteristics from sensing data and generate a control instruction. On the basis of a deep learning algorithm, an automatic intervention mechanism is designed, and when the control of the unmanned ship is interfered by the external environment, a control strategy can be automatically adjusted to maintain the accuracy and stability of the control. And a real-time monitoring system is established, and the control process of the unmanned ship is monitored and evaluated in real time. By collecting and analyzing feedback information of the unmanned ship, the performance of the automatic intervention mechanism is continuously adjusted and optimized. The problem caused by the lack of an effective automatic intervention mechanism of the unmanned ship is effectively solved, and the accuracy and stability of unmanned ship control are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned surface vessel (USV) control technology, and more particularly to a method for controlling an USV. Background Technology

[0002] As unmanned surface vessels (USVs) become more widely used, the precision and stability of their operation have become urgent issues to be addressed.

[0003] However, in some situations, manual operation is still necessary for the existing unmanned surface vessel (USV) control systems on the market, especially in complex or special mission environments. Operators can select manual control mode through the interactive panel to directly control the USV's forward, reverse, left and right turns.

[0004] However, the lack of an effective automatic intervention mechanism makes the operation of unmanned surface vessels susceptible to the influence of the external environment, and the accuracy and stability of operation cannot be guaranteed. Summary of the Invention

[0005] The purpose of this invention is to provide a method for controlling an unmanned surface vessel (USV), which solves the problem that existing USV control systems lack an effective automatic intervention mechanism, making the control of USVs susceptible to external environmental influences and unable to guarantee the accuracy and stability of control.

[0006] To achieve the above objectives, the present invention provides a method for controlling an unmanned surface vessel, comprising the following steps: Upgrade the sensing devices of unmanned surface vessels to acquire environmental perception data; Design and optimize deep learning algorithms to extract features from perceptual data and generate control commands; Design an automatic intervention mechanism based on deep learning algorithms; Establish a real-time monitoring system to monitor and evaluate the operation process of unmanned surface vessels in real time; The unmanned surface vessel was tested and verified, and the deep learning algorithm and automatic intervention mechanism were continuously optimized and adjusted based on the test results.

[0007] The step of upgrading the unmanned surface vessel's sensing devices to acquire environmental perception data also includes: Multi-sensor fusion technology is introduced to fuse data collected by different sensors.

[0008] The step of designing and optimizing a deep learning algorithm to extract features from perceptual data and generate control commands further includes: Collect and prepare sensory data; Select a deep learning model architecture based on task requirements and data characteristics; Design network layers and activation functions, use unsupervised learning methods for feature pre-training, and introduce attention mechanisms or feature fusion techniques to enhance the algorithm's ability to focus on key features; Design the output layer; Model training is performed using loss functions and optimization algorithms; The trained model is evaluated using a validation set to analyze its performance, and the model is then tuned based on the evaluation results.

[0009] The automatic intervention mechanism, based on deep learning algorithms, is designed, and the steps further include: Collect unmanned surface vessel (USV) operation data under various environments and conditions, label the data, and identify when automatic intervention is needed, as well as the goals and expected results of the intervention. The deep learning model is trained using labeled data, enabling it to identify scenarios requiring automatic intervention and predict corresponding control commands. Based on the model's predictions, design automatic intervention strategies; Integrate automatic intervention mechanisms into the control system of unmanned surface vessels.

[0010] The step of establishing a real-time monitoring system to monitor and evaluate the operation of the unmanned surface vessel in real time includes: Real-time acquisition of unmanned surface vessel status information, environmental perception data, and control commands; The received data is preprocessed, and features are extracted and patterns are recognized from the perceived data to assess the unmanned surface vessel's control status and environmental changes. The monitoring interface displays the unmanned surface vessel's position, heading, and speed information in real time. Based on the unmanned surface vessel's control data and environmental information, it can detect any abnormal conditions or potential risks in real time. The unmanned surface vessel's handling performance is evaluated in real time, and the evaluation results are fed back to the operator or control system. Establish a data storage mechanism to persistently save all data generated during the monitoring process.

[0011] This includes setting up the testing environment; Test the basic control functions of the unmanned surface vessel; Quantitatively evaluate the handling performance of unmanned surface vessels; Analyze the test data to identify potential errors or shortcomings in the algorithm and mechanism; Based on the test results, adjust the parameters, improve the structure, or update the model of the deep learning algorithm, and optimize the triggering conditions and intervention strategies of the automatic intervention mechanism.

[0012] This invention discloses a method for controlling an unmanned surface vessel (USV), upgrading the USV's perception devices by using higher-resolution cameras, more sensitive radar, and sonar equipment to acquire more accurate environmental data. Multi-sensor fusion technology is introduced to fuse data collected from different sensors, improving the reliability and completeness of perception information. Deep learning algorithms are designed and optimized to better extract features from perception data and generate control commands, specifically tailored to the characteristics of USV control. Transfer learning and reinforcement learning techniques are incorporated to enhance the algorithm's generalization ability and adaptability, enabling it to handle different environments and task requirements. Based on deep learning algorithms, an automatic intervention mechanism is designed to automatically adjust the control strategy to maintain accuracy and stability when USV control is disturbed by the external environment. Appropriate thresholds and trigger conditions are set based on USV status information and environmental data; the automatic intervention mechanism is activated when these conditions are met. A real-time monitoring system is established to monitor and evaluate the USV control process in real time. The performance of the automatic intervention mechanism is continuously adjusted and optimized by collecting and analyzing feedback information from the USV. By following the above methods and steps, the problem caused by the lack of an effective automatic intervention mechanism for unmanned surface vessels (USVs) can be effectively solved, thereby improving the accuracy and stability of USV control. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0014] Figure 1 This is a flowchart illustrating the operation method of the unmanned surface vessel of the present invention.

[0015] Figure 2 This invention presents a flowchart illustrating the steps involved in designing and optimizing a deep learning algorithm to extract features from perceptual data and generate control commands.

[0016] Figure 3 This is a flowchart illustrating the steps of designing an automatic intervention mechanism based on deep learning algorithms according to the present invention.

[0017] Figure 4 This invention provides a step-by-step diagram of establishing a real-time monitoring system to monitor and evaluate the operation process of unmanned surface vessels in real time.

[0018] Figure 5 This invention is a flowchart illustrating the steps of testing and verifying unmanned surface vessels, and continuously optimizing and adjusting deep learning algorithms and automatic intervention mechanisms based on test results. Detailed Implementation

[0019] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0020] Please see Figures 1 to 5 ,in, Figure 1 This is a flowchart illustrating the operation method of the unmanned surface vessel of the present invention. Figure 2 This invention presents a flowchart illustrating the steps involved in designing and optimizing a deep learning algorithm to extract features from perceptual data and generate control commands. Figure 3 This is a flowchart illustrating the steps of designing an automatic intervention mechanism based on deep learning algorithms according to the present invention. Figure 4 This invention provides a step-by-step diagram of establishing a real-time monitoring system to monitor and evaluate the operation process of unmanned surface vessels in real time. Figure 5 This invention involves testing and verifying unmanned surface vessels (USVs), and continuously optimizing and adjusting the deep learning algorithm and automatic intervention mechanism based on the test results. This invention also provides a method for controlling an USV, comprising the following steps: S100: Upgrade the unmanned surface vessel's sensing devices to acquire environmental perception data; Specifically, clearly define the types and accuracy requirements of environmental perception information needed by the unmanned surface vessel (USV) during missions. Analyze the performance and limitations of existing perception devices to determine which sensors need to be upgraded or added. Based on the requirements analysis, select suitable sensor types and models, including high-resolution cameras, radar, LiDAR, sonar, and infrared sensors. Evaluate the performance, reliability, cost, and integration difficulty of different sensors and procure them accordingly. Design sensor mounting brackets and fixing methods to ensure stable installation on the USV. Perform electrical connections for the sensors, including power lines and signal lines, to ensure proper operation and communication with the USV's control system. Configure sensor software parameters, such as sampling frequency, resolution, and range, to meet the USV's perception requirements. Calibrate the sensors to eliminate errors and biases, ensuring the accuracy of the perception data. If multiple sensors are upgraded, a data fusion algorithm needs to be designed to integrate and process data from different sensors to improve the integrity and reliability of the perception information. Use filtering algorithms and feature extraction techniques to preprocess the perception data, eliminating noise and interference and extracting useful feature information.

[0021] S200: Design and optimize deep learning algorithms to extract features from perceptual data and generate control commands; S201: Collect and prepare sensor data; S202: Select a deep learning model architecture based on task requirements and data characteristics; S203: Design network layers and activation functions, use unsupervised learning methods for feature pre-training, and introduce attention mechanisms or feature fusion techniques to enhance the algorithm's ability to focus on key features; S204: Design output layer; S205: Model training using loss functions and optimization algorithms; S206: Use the validation set to evaluate the trained model, analyze its performance, and optimize the model based on the evaluation results.

[0022] Specifically, clearly define the problem and the specific task the algorithm needs to solve, such as object detection, path planning, or control command generation. Collect and prepare a large amount of sensing data, including images, radar data, and sensor signals, for training and validating the algorithm. Select an appropriate deep learning model architecture based on task requirements and data characteristics, such as Convolutional Neural Networks (CNNs) for image processing and Recurrent Neural Networks (RNNs) for sequence data processing. Consider using a pre-trained model as a starting point to accelerate the training process and improve performance through transfer learning. Design appropriate network layers and activation functions to extract useful features from the sensing data. Use unsupervised learning methods (such as autoencoders) for feature pre-training to capture the inherent structure of the data. Introduce attention mechanisms or feature fusion techniques to enhance the algorithm's ability to focus on key features. Design the output layer to generate commands suitable for unmanned surface vessel control, such as heading angle and speed. Use regression models to directly predict control commands or use classification models to map sensing data to predefined control actions. Consider introducing reinforcement learning techniques to enable the algorithm to continuously optimize control commands through interaction with the environment. Train the model using appropriate loss functions and optimization algorithms, such as cross-entropy loss for classification tasks and mean squared error loss for regression tasks. Monitor performance metrics during training, such as accuracy, recall, or error control, and adjust model parameters or structure in a timely manner. Apply regularization techniques, batch normalization, or dropout to prevent overfitting and improve the model's generalization ability. Evaluate the trained model using a validation set and analyze its performance. Optimize the model based on the evaluation results, including adjusting hyperparameters such as network structure, learning rate, and batch size. Consider using ensemble learning or model fusion techniques to combine the advantages of multiple models and improve overall performance.

[0023] S300: Based on deep learning algorithms, an automatic intervention mechanism is designed; S301: Collect operational data of unmanned surface vessels under various environments and conditions, label the data, identify when automatic intervention is required, and the goals and expected results of the intervention; S302: Train a deep learning model using labeled data so that it can identify scenarios that require automatic intervention and predict the corresponding control commands; S303: Design automatic intervention strategies based on the model prediction results; S304: Integrate automatic intervention mechanisms into the control system of unmanned surface vessels.

[0024] Specifically, the goals and requirements of the automatic intervention mechanism should be clearly defined, such as identifying abnormal states, correcting control deviations, and responding to sudden environmental changes. Analyze potential interference factors encountered during unmanned surface vessel (USV) operation, such as wind, waves, currents, and obstacles, and identify scenarios requiring automatic intervention. Collect USV control data under various environmental and conditions, including data from normal and abnormal states. Label the data to clarify when automatic intervention is needed, as well as the intervention goals and expected results. Select a suitable deep learning model for processing time-series or image data, such as a recurrent neural network (RNN), long short-term memory network (LSTM), or convolutional neural network (CNN). Train the model using labeled data to enable it to identify scenarios requiring automatic intervention and predict corresponding control commands. Based on the model's predictions, design an automatic intervention strategy. This strategy includes adjusting the USV's heading, speed, and attitude to cope with external interference. Determine the timing and extent of intervention to ensure it is timely and effective without excessively affecting the USV's normal operation. Integrate the automatic intervention mechanism into the USV's control system, ensuring seamless integration with existing perception, decision-making, and execution systems. The integrated system was tested under different environmental and task conditions to verify the effectiveness and stability of the automatic intervention mechanism.

[0025] S400: Establish a real-time monitoring system to monitor and evaluate the operation of unmanned surface vessels in real time; S401: Real-time acquisition of unmanned surface vessel status information, environmental perception data, and control commands; S402: Preprocess the received data, extract features and recognize patterns from the perceived data, and assess the unmanned surface vessel's control status and environmental changes. S403: The unmanned surface vessel's position, heading, and speed information are displayed in real time through the monitoring interface; S404: Based on the unmanned surface vessel's control data and environmental information, detect possible abnormal states or potential risks in real time; S405: Performs real-time assessment of the unmanned surface vessel's handling performance and feeds back the assessment results to the operator or control system; S406: Establish a data storage mechanism to persistently save all data generated during the monitoring process.

[0026] Specifically, the overall architecture of the real-time monitoring system should be determined, including modules for data acquisition, transmission, processing, display, and storage. Appropriate hardware and software platforms should be selected to ensure stable and reliable system operation. Various sensors and control systems on the unmanned surface vessel (USV) should be integrated to collect real-time status information, environmental perception data, and control commands. A data transmission channel should be established to ensure that the collected data is transmitted to the monitoring center in real time and accurately. The received data should be preprocessed, including data cleaning, format conversion, and standardization. Deep learning algorithms or other data processing techniques should be used to extract features and perform pattern recognition on the perceived data to assess the USV's operational status and environmental changes. An intuitive and user-friendly monitoring interface should be designed to display key information such as the USV's position, heading, and speed in real time. Visual elements such as charts and animations should be integrated to help operators intuitively understand the USV's operation process and status changes. An anomaly detection algorithm should be designed to detect potential abnormal states or risks in real time based on the USV's operational data and environmental information. Once an anomaly is detected, the system should immediately trigger an early warning mechanism, alerting operators through sound, lights, etc., and taking necessary intervention measures. Set appropriate evaluation metrics, such as control precision, stability, and response speed, to conduct real-time assessments of the unmanned surface vessel's (USV) control performance. Feedback the evaluation results to operators or the control system to optimize USV control strategies or adjust deep learning algorithm parameters. Establish a data storage mechanism to persistently store all data generated during monitoring. Provide data backtracking functionality, allowing operators or researchers to access historical data for post-event analysis and improvement when needed.

[0027] S500: Test and verify unmanned surface vessels, and continuously optimize and adjust deep learning algorithms and automatic intervention mechanisms based on test results.

[0028] S501: Setting up the test environment; S502: Test the basic control functions of the unmanned surface vessel; S503: Quantitatively evaluate the handling performance of unmanned surface vessels; S504: Analyze test data to identify potential errors or deficiencies in the algorithm and mechanism; S505: Based on the test results, adjust the parameters, improve the structure, or update the model of the deep learning algorithm, and optimize the triggering conditions and intervention strategies of the automatic intervention mechanism.

[0029] Specifically, based on the unmanned surface vessel's (USV) control requirements and application scenarios, a suitable test environment should be built. This includes simulated experimental environments, closed water test fields, or open water test areas. The test environment should ensure it can simulate various real-world control conditions, including different weather conditions, water currents, and obstacles, to comprehensively evaluate the USV's performance. Basic USV control functions, such as start-up, navigation, turning, and stopping, should be tested to ensure all functions are normal. The processing capability and feature extraction effect of deep learning algorithms on perceived data should be verified, checking their ability to accurately identify environmental information. The response speed and accuracy of automatic intervention mechanisms in different scenarios should be tested to ensure timely and effective intervention in USV control. The USV's control performance should be quantitatively evaluated, including indicators such as stability, accuracy, and response speed. The impact of deep learning algorithms and automatic intervention mechanisms on USV performance should be analyzed to identify potential problems and areas for improvement. Test data should be carefully analyzed to identify potential errors or deficiencies in the algorithms and mechanisms. Based on the performance during actual control, the causes and locations of problems should be pinpointed, providing a basis for optimization and adjustments. Based on the test results, adjust the parameters, improve the structure, or update the model of the deep learning algorithm to enhance its ability to process perceived data and generate control commands. Optimize the triggering conditions and intervention strategies of the automatic intervention mechanism to more accurately determine when and how to intervene. Consider introducing new algorithms or technologies, such as reinforcement learning and transfer learning, to further improve the maneuverability of the unmanned surface vessel (USV). Redeploy the optimized deep learning algorithm and automatic intervention mechanism to the USV for a new round of testing and verification. Repeat the above steps, iteratively optimizing until the USV's performance meets the expected requirements. Record the process, results, and optimization measures of each test, forming a detailed test report and documentation. Summarize the experiences and lessons learned during the testing process to provide reference and guidance for future research and development.

[0030] By following the above methods and steps, the problem caused by the lack of an effective automatic intervention mechanism for unmanned surface vessels (USVs) can be effectively solved, thereby improving the accuracy and stability of USV control.

[0031] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A method for controlling an unmanned surface vessel, characterized in that, Includes the following steps: Upgrade the sensing devices of unmanned surface vessels to acquire environmental perception data; Design and optimize deep learning algorithms to extract features from perceptual data and generate control commands; Design an automatic intervention mechanism based on deep learning algorithms; Establish a real-time monitoring system to monitor and evaluate the operation process of unmanned surface vessels in real time; The unmanned surface vessel was tested and verified, and the deep learning algorithm and automatic intervention mechanism were continuously optimized and adjusted based on the test results.

2. The method for controlling an unmanned surface vessel as described in claim 1, characterized in that, Upgrading the unmanned surface vessel's sensing devices to acquire environmental perception data, the steps also include: Multi-sensor fusion technology is introduced to fuse data collected by different sensors.

3. The unmanned surface vessel control method as described in claim 2, characterized in that, The steps of designing and optimizing deep learning algorithms to extract features from perceptual data and generate control commands further include: Collect and prepare sensory data; Select a deep learning model architecture based on task requirements and data characteristics; Design network layers and activation functions, use unsupervised learning methods for feature pre-training, and introduce attention mechanisms or feature fusion techniques to enhance the algorithm's ability to focus on key features; Design the output layer; Model training is performed using loss functions and optimization algorithms; The trained model is evaluated using a validation set to analyze its performance, and the model is then tuned based on the evaluation results.

4. The unmanned surface vessel control method as described in claim 3, characterized in that, Based on deep learning algorithms, an automatic intervention mechanism is designed, the steps of which further include: Collect unmanned surface vessel (USV) operation data under various environments and conditions, label the data, and identify when automatic intervention is needed, as well as the goals and expected results of the intervention. The deep learning model is trained using labeled data, enabling it to identify scenarios requiring automatic intervention and predict corresponding control commands. Based on the model's predictions, design automatic intervention strategies; Integrate automatic intervention mechanisms into the control system of unmanned surface vessels.

5. The method for controlling an unmanned surface vessel as described in claim 4, characterized in that, Establish a real-time monitoring system to monitor and evaluate the operation process of the unmanned surface vessel in real time. The steps also include: Real-time acquisition of unmanned surface vessel status information, environmental perception data, and control commands; The received data is preprocessed, and features are extracted and patterns are recognized from the perceived data to assess the unmanned surface vessel's control status and environmental changes. The monitoring interface displays the unmanned surface vessel's position, heading, and speed information in real time. Based on the unmanned surface vessel's control data and environmental information, it can detect any abnormal conditions or potential risks in real time. The unmanned surface vessel's handling performance is evaluated in real time, and the evaluation results are fed back to the operator or control system. Establish a data storage mechanism to persistently save all data generated during the monitoring process.

6. The method for controlling an unmanned surface vessel as described in claim 5, characterized in that, Set up the test environment; Test the basic control functions of the unmanned surface vessel; Quantitatively evaluate the handling performance of unmanned surface vessels; Analyze the test data to identify potential errors or shortcomings in the algorithm and mechanism; Based on the test results, adjust the parameters, improve the structure, or update the model of the deep learning algorithm, and optimize the triggering conditions and intervention strategies of the automatic intervention mechanism.

Images