Test method, test device, electronic equipment, readable storage medium and chip

Abstract

The invention provides a test method, a test device, electronic equipment, a readable storage medium and a chip, the test method is used for at least one unmanned ship, and the test method comprises the following steps: obtaining test environment parameters and a simulation scene corresponding to the test environment parameters, and collecting motion data of the at least one unmanned ship in the simulation scene through a sensor; determining at least one piece of annotation information associated with the motion data, storing the corresponding annotation information and the motion data in a database in a preset format, obtaining a training algorithm model, determining test state information based on the training algorithm model and the corresponding annotation information and the motion data, and according to the test state information, the test environment parameters and the motion data, obtaining a test result. Evaluation parameters corresponding to the at least one unmanned ship are determined.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned boats, and in particular, to a test method, a test device, an electronic device, a readable storage medium, and a chip. Background Art

[0002] With the rapid development and wide application of unmanned boat technology, it is particularly important to create an intelligent learning environment for unmanned boats, cultivate the autonomous decision-making and adaptation capabilities of unmanned boat systems, enable them to cope with complex and dynamic marine environments, and perform various tasks. In the existing technical solutions, by constructing an intelligent simulation system for surface unmanned boats and establishing a virtual environment, the testing of unmanned boats in a virtual scenario is realized. However, during the testing process, the obtained test data lacks annotations, and the analysis results are too single, resulting in many information blanks in the test report, which affects the test judgment. Summary of the Invention

[0003] In view of this, the present invention aims to solve the problem of blank information in unmanned boat test data.

[0004] Specifically, the present invention is implemented through the following technical solutions:

[0005] The first aspect of the present invention provides a test method.

[0006] The second aspect of the present invention provides a test device.

[0007] The third aspect of the present invention provides an electronic device.

[0008] The fourth aspect of the present invention provides a readable storage medium.

[0009] The fifth aspect of the present invention provides a chip.

[0010] The test method for an unmanned boat provided by the present invention is used for at least one unmanned boat. The test method includes: obtaining test environment parameters and a simulation scenario corresponding to the test environment parameters; collecting motion data of at least one unmanned boat in the simulation scenario through a sensor; determining at least one piece of annotation information associated with the motion data; storing the corresponding annotation information and motion data in a database in a preset format; obtaining a training algorithm model; determining test status information based on the training algorithm model and the corresponding annotation information and motion data; and determining evaluation parameters corresponding to at least one unmanned boat according to the test status information, test environment parameters, and motion data.

[0011] In some technical solutions, optionally, obtaining test environment parameters and a simulation scenario corresponding to the test environment parameters specifically includes: responding to a creation instruction to obtain test environment parameters; determining environment information and weather information according to the test environment parameters; and creating a simulation scenario according to the environment information and weather information.

[0012] In some technical solutions, optionally, obtaining test environment parameters and a simulation scenario corresponding to the test environment parameters specifically includes: determining, in response to an import instruction, test environment parameters related to environment information and weather information stored in a data source; determining a simulation scenario according to the test environment parameters; or determining, in response to an import instruction, model information corresponding to the test environment parameters, where the model information includes environment information and weather information; determining a simulation scenario according to the model information.

[0013] In some technical solutions, optionally, obtaining test environment parameters and a simulation scenario corresponding to the test environment parameters specifically includes: determining, in response to a search instruction, search information related to the test environment parameters; determining a simulation scenario corresponding to the search information in a historical scenario database.

[0014] In some technical solutions, optionally, before collecting motion data of at least one unmanned boat in a simulation scenario through sensors, it further includes: receiving an adjustment instruction corresponding to the simulation scenario; adjusting the test environment parameters according to the adjustment instruction, and establishing a simulation scenario according to the adjusted test environment parameters.

[0015] In some technical solutions, optionally, collecting motion data of at least one unmanned boat in a simulation scenario through sensors specifically includes: determining, through at least one sensor disposed on the unmanned boat, environment data of the environment where the unmanned boat is located; determining, through at least one positioning sensor disposed on the unmanned boat, navigation data of the unmanned boat, where the navigation data includes position, speed, acceleration, heading, and turning rate; determining, through an inertial sensor disposed on the unmanned boat, attitude data of the unmanned boat, where the attitude data includes pitch, roll, yaw attitude angles, and angular velocity.

[0016] In some technical solutions, optionally, storing corresponding annotation information and motion data in a database in a preset format specifically includes: obtaining a preset format for storing data and an encryption algorithm for encrypting data in the database; preprocessing the associated annotation information and motion data to generate preprocessed data; encrypting the preprocessed data through the encryption algorithm to generate data to be stored; storing the data to be stored in the database.

[0017] In some technical solutions, optionally, obtaining a training algorithm model specifically includes: determining an algorithm library storing multiple preset algorithm models; obtaining configuration parameters corresponding to at least one preset algorithm model; determining a training algorithm model according to the configuration parameters.

[0018] In some technical solutions, optionally, based on the training algorithm model, the corresponding annotation information, and the motion data, the test state information is determined, specifically including: based on the training algorithm model, the corresponding annotation information and the motion data are input, and the visualization tool is used to output simulation data, where the simulation data includes simulation environment information and object position information.

[0019] In some technical solutions, optionally, according to the test state information, the test environment parameters, and the motion data, the evaluation parameters corresponding to at least one unmanned boat are determined, specifically including: when the number of unmanned boats is one, the trajectory information of the unmanned boat is determined according to the simulation data; when the number of unmanned boats is multiple, the boat group information of the multiple unmanned boats is determined according to the simulation data, and the boat group information includes cooperation information, formation information, and system information.

[0020] The second aspect of the present invention provides a test device for at least one unmanned boat. The test device includes: an acquisition module for acquiring test environment parameters and a simulation scenario corresponding to the test environment parameters; a collection module for collecting motion data of at least one unmanned boat in the simulation scenario through a sensor; an annotation module for determining at least one annotation information associated with the motion data; a storage module for storing the corresponding annotation information and the motion data in a preset format in a database; a model acquisition module for acquiring a training algorithm model; a test module for determining test state information based on the training algorithm model and the corresponding annotation information and motion data; and an evaluation module for determining evaluation parameters corresponding to at least one unmanned boat according to the test state information, the test environment parameters, and the motion data.

[0021] The third aspect of the present invention provides an electronic device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of the test method according to the first aspect of the present invention are implemented.

[0022] The fourth aspect of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by the processor, the steps of the test method according to the first aspect of the present invention are implemented.

[0023] The fifth aspect of the present invention provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps of the test method according to the first aspect of the present invention.

[0024] The technical solutions provided by the present invention at least bring the following beneficial effects:

[0025] The present invention provides a method for testing an unmanned boat that can be marked. By marking the historical motion data collected by the unmanned boat and bringing the marked data into the operation to output the result, it fills in the blanks of some data in the analysis report, makes the analysis data complete, and improves the accuracy and pertinence of the judgment of the test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or related technologies. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0028] Figure 1 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0029] Figure 2 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0030] Figure 3 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0031] Figure 4 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0032] Figure 5 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0033] Figure 6 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0034] Figure 7 It is a schematic flowchart of the testing method for the unmanned boat provided by the embodiment of the present invention;

[0035] Figure 8 It is a schematic structural diagram of the testing device for the unmanned boat provided by the embodiment of the present invention;

[0036] Figure 9 It is a schematic structural diagram of the electronic device provided by the embodiment of the present invention;

[0037] Figure 10 It is a typical ability learning flowchart provided by the embodiment of the present invention;

[0038] Figure 11Schematic diagram of collecting historical data at the end of the boat provided by the embodiment of the present invention;

[0039] Figure 12 Schematic diagram of the data annotation tool provided by the embodiment of the present invention;

[0040] Figure 13 Schematic diagram of managing the algorithm learning environment provided by the embodiment of the present invention;

[0041] Figure 14 Schematic diagram of algorithm training provided by the embodiment of the present invention;

[0042] Figure 15 Schematic diagram of model verification provided by the embodiment of the present invention.

[0043] Among them, Figure 8 and Figure 9 The corresponding relationship between the part names and labels is as follows:

[0044] 200 test device, 201 acquisition module, 202 acquisition module, 203 annotation module, 204 storage module, 205 model acquisition module, 206 test module, 207 evaluation module, 300 electronic device, 302 processor, 304 memory. Specific implementation manners

[0045] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Refer to Figure 1 and Figure 10 , the first aspect of the present invention provides a test method for an unmanned boat, which is used for at least one unmanned boat, and the test method includes:

[0047] S102: Obtain test environment parameters and a simulation scenario corresponding to the test environment parameters;

[0048] S104: Collect motion data of at least one unmanned boat in the simulation scenario through sensors;

[0049] S106: Determine at least one piece of annotation information associated with the motion data;

[0050] S108: Store the corresponding annotation information and motion data in a database in a preset format;

[0051] S110: Obtain a training algorithm model;

[0052] S112: Determine the test status information based on the training algorithm model, the corresponding annotation information, and the motion data.

[0053] S114: Determine the evaluation parameters corresponding to at least one unmanned boat according to the test status information, the test environment parameters, and the motion data.

[0054] For the unmanned boat test method provided in this embodiment, it is first necessary to obtain the test environment parameters, and construct a test environment model based on the obtained test environment parameters, that is, a simulation scenario for the unmanned boat test can be built through the test environment parameters, so that the unmanned boat can conduct targeted test exercises in the assumed test scenario. During the test process, one or more unmanned boats can collect motion data in the simulation scenario through the settings of their own sensors. The motion data includes: navigation information data, attitude information data, and sensor data. By collecting these motion data, the unmanned boat can better adapt to the surrounding environment and its own motion conditions, which is essential for the unmanned boat to engage in maritime search and rescue tasks in the future, and helps the unmanned boat cope with the complex and changeable maritime environment. Further, one or more motion data collected by the unmanned boat in the simulation scenario are annotated. The annotation method can be manual annotation, semi-automatic annotation, or automatic annotation. Add tag annotations to one or more historical data collected by the unmanned boat, such as time stamps, environmental conditions, task types, etc., and add metadata to each data point or timestamp that needs to be marked. These tag annotations provide the associated information of the original data, can provide corresponding explanations and classifications for the data, and help associate the data with specific scenarios and tasks. Then, the motion data collected by the unmanned boat and the corresponding annotation information are stored in the database in a preset format for algorithm training and evaluation. Through the system setting of the algorithm library before training, a variety of different intelligent algorithm options are provided. According to different test schemes, the required algorithm model is selected. Then, based on the selected training algorithm model, data processing is performed on the motion data and the corresponding annotation information in the database. Through the calculation of the algorithm model, the test status information is determined, and the execution degree of the unmanned boat on the training content and training tasks in the test environment is understood. By comparing and analyzing the training status information, motion data information, and experimental environment parameters of the unmanned boat, the corresponding evaluation parameters of the unmanned boat or the corresponding evaluation parameters of the unmanned boat group are determined for evaluating the results of this test. Specifically, the test process and test results are visually presented to help improve the performance of the evaluation system, improve the algorithm, verify the model, and optimize the operation, which can continuously improve the behavior, performance, and safety of the boat group, and this helps to ensure the reliability and efficiency of the surface unmanned boat in different tasks and environments.

[0055] It can be understood that the unmanned boat test method of this application is to first obtain test environment parameters. The test environment parameters can build a simulation environment for unmanned boat tests, providing a simulation platform for unmanned boat tests. Through the collection of motion data by the sensors of the unmanned boat itself, the unmanned boat can better understand the surrounding environment and its own situation, and label the collected motion data. The labeled data is helpful for subsequent calculations and analyses. Subsequently, it is stored in a database in a preset format for subsequent training calculations. By determining the algorithm model, an algorithm channel corresponding to the test content is selected. Further, by processing and calculating the motion data and the corresponding annotation information in the database, test status information is obtained. Finally, according to the test status information, test environment parameters, and motion data, the evaluation parameters corresponding to the unmanned boat or unmanned boat group are determined.

[0056] Briefly speaking, before the test starts, first formulate an unmanned boat sea test outline, use the test integrated control platform to compile a scenario-based test plan, conduct targeted tests on the unmanned boat, and at the same time obtain test environment parameters. Use the test environment parameters to construct a simulation environment to prepare for the unmanned boat sea test. After all preparations are ready, start the test. The unmanned boat collects motion data in the simulation environment, labels the motion data collected by the unmanned boat in the simulation environment, and stores the labeled annotation information together with the motion data in the database. Subsequently, they participate in model training together to obtain an algorithm model, and it is allowed to visually monitor the process of algorithm model training, display test status information, and determine the evaluation parameters corresponding to at least one unmanned boat according to the test status information, test environment parameters, and motion data. Generate an evaluation report. Using such a design method helps to improve the accuracy and integrity of test results, enhances the autonomous learning ability of unmanned boats, helps to organize and classify the data collected during the operation of multiple unmanned boats, helps these autonomous systems to better adapt to changing tasks and environmental conditions, improves their performance and usability, reduces operating costs, and enhances the application potential in various application fields. This has important value for tasks such as marine scientific research, resource exploration, search and rescue, and environmental monitoring.

[0057] As Figure 2 shown, in some embodiments, optionally, obtaining the test environment parameters and the simulation scenario corresponding to the test environment parameters specifically includes:

[0058] S1021: Respond to the creation instruction to obtain the test environment parameters;

[0059] S1022: Determine the environmental information and weather information according to the test environment parameters;

[0060] S1023: Create a simulation scenario according to the environmental information and weather information.

[0061] In this embodiment, by responding to a creation instruction, test environment parameters are obtained. That is, the user can, according to the experimental content and test plan, use the creation function to set the environment parameters by himself / herself, and determine the environmental information and weather information according to the set test environment parameters. Among them, the environmental information includes: water surface information such as oceans, lakes, and rivers, including: geographical information such as coastlines, ports, and islands, and also includes: obstacle information such as enemy ships, buoys, and warships. The weather information includes: different marine meteorological conditions such as wind speed, tides, and wave height. By creating the test environmental information and weather information, a simulation scenario for unmanned boats and unmanned boat groups is built.

[0062] It can be understood that the user can, according to the needs of the unmanned boat experimental content, envision and create different water area simulation environments. By setting the weather and environment in the simulation scenario, the ability of the unmanned boat to avoid obstacles and cooperate can be tested, and the difficulty of the test can be increased to adapt to the complex and changeable marine environment.

[0063] As Figure 3 shown, in some embodiments, optionally, obtaining the test environment parameters and the simulation scenario corresponding to the test environment parameters specifically includes:

[0064] S1024: Respond to the import instruction to determine the test environment parameters related to the environmental information and weather information stored in the data source; or respond to the import instruction to determine the model information corresponding to the test environment parameters, where the model information includes environmental information and weather information;

[0065] S1025: Determine the simulation scenario according to the test environment parameters; or determine the simulation scenario according to the model information.

[0066] In this embodiment, by responding to the import instruction of the system, the test environment parameters related to the environmental information and weather information stored in the data source are determined, and the simulation scenario is determined according to the test environment parameters related to the environmental information and weather information. Or respond to the import instruction to determine the environmental model information corresponding to the test environment and the model information of the weather, and then determine the simulation scenario according to the model information. It can be understood that the user can use two methods to determine the simulation scenario by using the import instruction of the system. One is data import, where the weather parameter information and environmental parameter information stored externally are imported into the simulation system to generate a simulation scenario. For example, the user is allowed to import map data, weather data, bathymetric data, etc. of the real-world water area from an external data source to more realistically simulate the learning environment. The other is direct model import, where the environment model and weather model for simulation generated from the already processed data are imported into the simulation system to build a simulation scenario. For example: support the import of existing unmanned surface vehicle models, environment models, physical engines, or simulation tools, without the need to process the test environment parameters again for use in the learning environment.

[0067] By responding to the import instruction and adopting various import methods of data models, the creation of the simulation scenario becomes multi-faceted, providing multiple construction channels. Using the existing data to create the simulation scenario can greatly shorten the pre-preparation time of the simulation experiment, improve the utilization rate of the model data, and can be reused only by creating it once. It can be used for repeated practice of the model, and can also achieve the maximization of the benefits of the model data through saving and sharing, which helps to improve the overall test efficiency during the unmanned boat test process.

[0068] As Figure 4 shown, in some embodiments, optionally, obtaining the test environment parameters and the simulation scenario corresponding to the test environment parameters specifically includes:

[0069] S1026: Determine the search information related to the test environment parameters in response to the search instruction;

[0070] S1027: Determine the simulation scenario corresponding to the search information in the historical scenario database.

[0071] In this embodiment, by responding to the search instruction of the simulation system, the search information related to the current test environment parameters is determined. These search information helps the subsequent retrieval of the existing simulation scenario data, greatly saving the query time. After determining the relevant search information, a query is performed in the historical repository set by the simulation system, that is, by retrieving the scenario model established for the experimental environment parameters created and imported in the test system in the past, and retrieving the scenario model corresponding to the search information. Generally speaking, by responding to the search instruction, it is convenient for users to quickly retrieve the used simulation scenario in the established historical scenario library, without the need to create the simulation scenario, saving the pre-preparation time of the test. Through the function that supports the storage of the simulation scenario, the utilization rate of the simulation scenario can be improved, which helps the repeated training of the unmanned boat test and makes the test results of the unmanned boat more persuasive. In addition, the available query method can facilitate the test personnel to record the past test process and provide a channel for memory.

[0072] It should be added that the search channels can be diverse. Specifically, the search information is defined as environmental search: allowing users to query the existing learning environment according to different criteria, such as geographical location, scenario type, difficulty level, etc.; it can also be designed as environmental browsing, providing a visual environmental browsing interface, allowing users to quickly browse and select the learning environment; it can also be queried through the historical record. Record the user's query history so that they can easily retrieve the previously used environment.

[0073] As Figure 5 shown, in some embodiments, optionally, before collecting the motion data of at least one unmanned boat in the simulation scenario through the sensor, it further includes:

[0074] S1032: Receive an adjustment instruction corresponding to the simulation scenario;

[0075] S1034: Adjust the test environment parameters according to the adjustment instruction, and establish a simulation scenario based on the adjusted test environment parameters.

[0076] In this embodiment, before the unmanned boat collects its own motion data in the simulation scenario through sensors, that is, after the simulation scenario is set up by means of creation, import, and query, it also meets the requirement of changing the data of the unmanned boat test simulation scenario, which can be temporarily changed according to the scenario plan or supplemented at any time to adjust the test plan, modify the experimental environment parameters, and establish a scenario model according to the final test environment parameters of the last adjustment, making the establishment of the scenario model flexible and capable of responding to changes in the test plan in a timely manner. For the above modifications to the test environment parameters, the following methods can be used to achieve them:

[0077] Environment editing: Allow users to edit and customize in the existing learning environment, including adding, deleting, or modifying geographical features, obstacles, targets, etc.

[0078] Parameter adjustment: Allow users to adjust the environment parameters at any time to change the ocean and weather conditions, and adjust the difficulty of the learning environment.

[0079] Version control: Provide environment version control so that users can save different versions of the learning environment and restore to a specific version when needed.

[0080] In some embodiments, optionally, collect the motion data of at least one unmanned boat in the simulation scenario through sensors, specifically including: determining the environmental data of the environment where the unmanned boat is located through at least one sensor provided on the unmanned boat; determining the navigation data of the unmanned boat through at least one positioning sensor provided on the unmanned boat, and the navigation data includes position, speed, acceleration, heading, and turning rate; determining the attitude data of the unmanned boat through the inertial sensor provided on the unmanned boat, and the attitude data includes pitch, roll, yaw attitude angles, and angular velocity.

[0081] In this embodiment, after the simulation scenario is created, start the simulation test on the unmanned boat or the unmanned boat group. The unmanned boat collects its own motion data in the simulation scenario through sensors. Among them, collect the environmental data where the unmanned boat is located through one or more sensors provided on the unmanned boat, such as underwater sonar data, camera images, water temperature, water quality, meteorological conditions, etc. These data are crucial for environmental perception and task execution, and help to understand the data collection ability of the unmanned boat for the surrounding environmental data, so as to provide data support for subsequent obstacle avoidance of the unmanned boat and complex sea weather training.

[0082] Navigation information data during the movement of the unmanned boat is collected by one or more positioning sensors on the unmanned boat. These data include: the position coordinates (longitude, latitude), speed, acceleration, heading, turning rate, etc. of the unmanned boat. The navigation information data is crucial for understanding the navigation performance and path planning of the unmanned boat, helps to complete the autonomous navigation of the unmanned boat, improves the route navigation planning ability, and realizes complex maritime tasks. Specifically, these data can be collected by the GPS and inertial navigation system on the unmanned boat.

[0083] Attitude data of the unmanned boat is collected by one or more inertial sensors on the unmanned boat. The attitude information data includes the pitch, roll, and yaw attitude angles of the unmanned boat, as well as the relevant angular velocity data. These data are very important for understanding the stability and handling performance of the unmanned boat, can be used to detect the stability during the execution of the unmanned boat task, and the operability when driving in bad maritime weather conditions, and improve the ability of the unmanned boat to autonomously adjust in complex situations. Specifically, it is usually collected by an Inertial Measurement Unit (IMU) or a gyroscope.

[0084] As Figure 6 shown, in some embodiments, optionally, the corresponding annotation information and motion data are stored in the database in a preset format, specifically including:

[0085] S1082: Obtain the preset format for storing data and the encryption algorithm for encrypting data in the database;

[0086] S1084: Preprocess the associated annotation information and motion data to generate preprocessed data;

[0087] S1086: Encrypt the preprocessed data through the encryption algorithm to generate data to be stored;

[0088] S1088: Store the data to be stored in the database.

[0089] In this embodiment, before storing the motion data and the corresponding annotation information in the database, it is necessary to first obtain the storage format preset in the database for storing data, store the data that meets the storage format requirements, and do not store the data that does not meet the storage format requirements. At the same time as determining the preset storage format, determine the encryption algorithm, that is, used to encrypt the stored data set to ensure the confidentiality and security of the data and prevent data leakage or loss.

[0090] First, the system preprocesses the motion data collected by the unmanned boat that needs to be stored and the annotation information associated with the motion data to make it meet the storage requirements and generate preprocessed data. The preprocessing data includes cleaning and preprocessing the original data, specifically including removing outliers, filling in missing data, aligning timestamps, etc., to prepare the data for subsequent storage and encryption. Then, the data processed by the preprocessing is encrypted through an encryption algorithm to form the data to be stored. The data in the database is further processed, and finally, the generated data to be stored is stored in the database, and the data in the database is encrypted and protected. Considering the possibility of sensitive data, appropriate data encryption and access control measures are taken to protect the data from unauthorized access and leakage.

[0091] In the above process, the types of annotations can be diverse. By setting different tag types by the user, the association of annotations with the original motion information is realized. Specifically, it can include: classification tags, which divide the data into different categories and types. For example, the animals in the image are marked as dogs, cats, birds, etc.; bounding box tags, which mark the target position and boundary position in the image, including information such as the coordinates, width, and height of the target; semantic segmentation tags, which specifically assign each pixel in the image to different semantic categories. For example, areas such as roads, vehicles, and pedestrians are segmented; instance segmentation tags, which are similar to semantic segmentation but distinguish different instances, that is, separate different objects in the same category; timestamp and event tags. For time series data, the tags include timestamps and tags related to specific events or states, such as sensor readings, action categories, etc.

[0092] The annotation method can be manual annotation. Manual annotation is the most common method, which involves manual observation and marking of data. This can be done through graphical interface tools, text editors, or custom software; semi-automatic annotation. Semi-automatic annotation tools combine manual annotation and automatic processing. For example, computer vision algorithms are used to automatically generate initial tags, and then manually reviewed and corrected; automatic annotation. For some tasks, machine learning or deep learning models can be used for automatic annotation. This is applicable to large-scale data sets, but high-quality training data is required to build the model.

[0093] The above annotation process is specifically described as follows. Before starting the annotation, a data set needs to be prepared, including original data, images, texts, sensor readings, etc. Then, an annotation strategy is defined, including which data needs to be annotated, how to annotate, whether there are multiple annotators, etc. For each data point, the annotator annotates according to the annotation strategy and associates the relevant information with the data. Quality control measures are implemented, such as double annotation, annotation consistency check, and error repair, to ensure the accuracy of the annotation. The annotated data is usually stored in a standard format for subsequent training and evaluation of machine learning models.

[0094] As Figure 7 shown, in some embodiments, optionally, obtaining a training algorithm model specifically includes:

[0095] S1102: Determine an algorithm library storing multiple preset algorithm models;

[0096] S1104: Obtain configuration parameters corresponding to at least one preset algorithm model;

[0097] S1106: Determine the training algorithm model according to the configuration parameters.

[0098] In this embodiment, first, an algorithm library is set up. By setting up the algorithm library, the algorithm library stores multiple preset algorithm models. These preset algorithm models can provide multiple different intelligent algorithm options for the calculation of the training program, including deep learning models, reinforcement learning algorithms, path planning algorithms, etc. Then, obtain at least one or more configuration parameters corresponding to the algorithm model, allowing the user to configure the parameters of the algorithm according to the research needs and the characteristics of the learning environment to optimize the performance. Finally, the training algorithm model finally used for the test scheme is determined by the configuration parameters.

[0099] In some embodiments, optionally, based on the training algorithm model and the corresponding annotation information and motion data, determine the test status information, specifically including: Based on the training algorithm model, input the corresponding annotation information and motion data, and use a visualization tool to output simulation data, where the simulation data includes simulation environment information and object position information.

[0100] In this embodiment, model visualization is a technology that presents the internal structure, parameters, training process, and prediction results of machine learning and deep learning models in a graphical or visual way. It is very important for understanding the model, diagnosing problems, optimizing performance, and sharing the model with stakeholders. In the calculation mode based on the training algorithm model, the motion data and the corresponding annotation information in the database are input into the algorithm model, and then the simulation data is output through a visualization tool, including simulation environment information, such as weather information such as wind force, sea waves, and weather, and object position information, such as the positions of civil ships, buoys, and other obstacles.

[0101] Among them, the visualization tool can achieve model structure visualization, including network structure diagrams: For deep learning models, the structure of the neural network can be visualized, including the connections, weights, and activation functions of each layer. This helps to understand the complexity and hierarchical structure of the model. Model architecture: Show the overall architecture of the model, including the input layer, hidden layer, output layer, and their connections. This helps to view the data flow and information transfer of the model.

[0102] Visualization of parameters, specifically, weight heatmap: The weights of the model can be visualized through a heatmap to help identify which weights have the greatest impact on the model's output; activation visualization: The activation values of each layer can be visualized to understand how the model processes the input data and to help detect potential problems.

[0103] Visualization of the training process, including: Training loss curve: The curve of the loss function during the training process can be plotted to evaluate the training progress and convergence of the model; learning rate curve: The change in the learning rate can be monitored to ensure the stability of the model during training.

[0104] In some embodiments, optionally, according to the trial status information, trial environment parameters, and motion data, evaluation parameters corresponding to at least one unmanned boat are determined, specifically including: When the number of unmanned boats is one, the trajectory information of the unmanned boat is determined according to the simulation data; when the number of unmanned boats is multiple, the boat group information of multiple unmanned boats is determined according to the simulation data, and the boat group information includes cooperation information, formation information, and system information.

[0105] In this embodiment, through the analysis and evaluation of the trial status information, trial environment parameters, and motion data, it is determined that when the number of unmanned boats is one, according to the calculation and analysis of the simulation data, the information on the movement trajectory of the unmanned boat can be determined, and relevant data such as the position and speed during its movement can be obtained; when the number of unmanned boats is multiple, according to the calculation and analysis of the simulation data, the boat group movement information of multiple unmanned boats is obtained, specifically including: calculating the distance, speed, and acceleration between each boat in the cluster to evaluate the cooperation and collision avoidance capabilities; evaluating usage metrics such as formation error and alignment to evaluate the quality of the formation; determining the task completion time, resource utilization rate, energy efficiency, etc., for evaluating the overall performance of the surface unmanned boat system.

[0106] Data analysis and evaluation provide key information for the design, optimization, and decision-making of the surface unmanned boat system. By comprehensively using sensor data, simulation platform data, and machine learning techniques, the behavior, performance, and safety of the boat group can be continuously improved. This helps to ensure the reliability and efficiency of the surface unmanned boat under different tasks and environments.

[0107] Such as Figure 8As shown in the figure, the second aspect of the present invention provides a test device 200 for an unmanned boat, which is used for at least one unmanned boat. The test device includes: an acquisition module 201, configured to acquire test environment parameters and a simulation scenario corresponding to the test environment parameters; a collection module 202, configured to collect motion data of at least one unmanned boat in the simulation scenario through sensors; a labeling module 203, configured to determine at least one piece of labeling information associated with the motion data; a storage module 204, configured to store the corresponding labeling information and motion data in a preset format in a database; a model acquisition module 205, configured to acquire a training algorithm model; a test module 206, configured to determine test status information based on the training algorithm model and the corresponding labeling information and motion data; and an evaluation module 207, configured to determine evaluation parameters corresponding to at least one unmanned boat according to the test status information, the test environment parameters, and the motion data.

[0108] In this embodiment, the present invention provides a test device 200 for an unmanned boat, including an acquisition module 201, a collection module 202, a labeling module 203, a storage module 204, a model acquisition module 205, a test module 206, and an evaluation module 207. Specifically, the acquisition module 201 is configured to acquire test environment parameters and a simulation scenario corresponding to the test environment parameters; the collection module 202 is configured to collect motion data of at least one unmanned boat in the simulation scenario through sensors; the labeling module 203 is configured to determine at least one piece of labeling information associated with the motion data; the storage module 204 is configured to store the corresponding labeling information and motion data in a preset format in a database; the model acquisition module 205 is configured to acquire a training algorithm model; the test module 206 is configured to determine test status information based on the training algorithm model and the corresponding labeling information and motion data; and the evaluation module 207 is configured to determine evaluation parameters corresponding to at least one unmanned boat according to the test status information, the test environment parameters, and the motion data.

[0109] In some embodiments, optionally, the test device is further configured to: respond to a creation instruction to acquire test environment parameters; determine environment information and weather information according to the test environment parameters; and create a simulation scenario according to the environment information and the weather information.

[0110] In some embodiments, optionally, the test device is further configured to: respond to an import instruction to determine test environment parameters stored in a data source and related to environment information and weather information; determine a simulation scenario according to the test environment parameters; or respond to an import instruction to determine model information corresponding to the test environment parameters, where the model information includes environment information and weather information; and determine a simulation scenario according to the model information.

[0111] In some embodiments, optionally, the testing device is further configured to: determine search information related to the test environment parameters in response to a search instruction; and determine a simulation scenario corresponding to the search information in the historical scenario database.

[0112] In some embodiments, optionally, the testing device is further configured to: receive an adjustment instruction corresponding to the simulation scenario; adjust the test environment parameters according to the adjustment instruction, and establish a simulation scenario according to the adjusted test environment parameters.

[0113] In some embodiments, optionally, the testing device is further configured to: determine environmental data of the environment where the unmanned boat is located through at least one sensor disposed on the unmanned boat; determine navigation data of the unmanned boat through at least one positioning sensor disposed on the unmanned boat, where the navigation data includes position, speed, acceleration, heading, and turning rate; and determine attitude data of the unmanned boat through an inertial sensor disposed on the unmanned boat, where the attitude data includes pitch, roll, yaw attitude angles, and angular velocity.

[0114] In some embodiments, optionally, the testing device is further configured to: obtain a preset format for storing data and an encryption algorithm for encrypting data in the database; preprocess the associated annotation information and motion data to generate preprocessed data; encrypt the preprocessed data through the encryption algorithm to generate data to be stored; and store the data to be stored in the database.

[0115] In some embodiments, optionally, the testing device is further configured to: determine an algorithm library storing multiple preset algorithm models; obtain configuration parameters corresponding to at least one preset algorithm model; and determine a training algorithm model according to the configuration parameters.

[0116] In some embodiments, optionally, the testing device is further configured to: based on the training algorithm model, input corresponding annotation information and motion data, and output simulation data using a visualization tool, where the simulation data includes simulation environment information and object position information.

[0117] In some embodiments, optionally, the testing device is further configured to: when the number of unmanned boats is one, determine the trajectory information of the unmanned boat according to the simulation data; and when the number of unmanned boats is multiple, determine the boat group information of multiple unmanned boats according to the simulation data, where the boat group information includes cooperation information, formation information, and system information.

[0118] Through the above unmanned boat testing device 200, a customizable and standardized test environment is constructed, enabling repeated testing of the boat group, synchronously collecting relevant data, and annotating, storing, calculating, and analyzing the collected data, thereby realizing full-dimensional testing and quantitative evaluation of the boat group and learning of typical capabilities. This has important value for tasks such as marine scientific research, resource exploration, search and rescue, and environmental monitoring.

[0119] Such as Figure 9As shown in the figure, the third aspect of the present invention provides an electronic device 300, which includes a processor 302, a memory 304, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, it implements the steps of the test method according to the first aspect of the present invention and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0120] It should be noted that the electronic devices in the embodiments of the present application include the above-mentioned electronic devices and non-electronic devices.

[0121] The fourth aspect of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by the processor, it implements the steps of the test method according to the first aspect of the present invention and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0122] The method can be implemented in various different ways according to specific features and / or example applications. For example, these methods can be implemented by a combination of hardware, firmware, and / or software. For example, in a hardware implementation, the processor can be implemented in one or more application-specific integrated circuits, digital signal processors, digital signal processing devices, programmable logic devices, field programmable gate arrays, controllers, microcontrollers, microprocessors, electronic devices, other device units for performing the above functions, and / or combinations thereof.

[0123] A computer-readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. A computer-readable storage medium can be an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above devices, but is not limited thereto. A non-exhaustive list of more specific examples of a computer-readable storage medium includes: portable computer floppy disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories, static random access memories, portable compact disc read-only memories, digital versatile disks, memory cards, floppy disks, encoding mechanical devices (such as punched cards or grooves with raised structures recording instructions), and any suitable combination of the above devices. The computer-readable storage medium used herein should not be construed as a signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagated through waveguides or other transmission media, or electrical signals transmitted through wires.

[0124] Among them, the processor is the processor in the electronic device in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0125] The fifth aspect of the present invention provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is configured to run programs or instructions to implement the steps of the testing method as in the first aspect of the present invention, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0126] It should be understood that the chip mentioned in the embodiments of the present application may also be referred to as a system-on-chip, system chip, chip system, or system-on-a-chip, etc.

[0127] In a specific embodiment, a customizable standardized testing environment is specifically constructed to achieve repeated testing of a fleet of boats, synchronously collect relevant data, achieve full-dimensional testing and quantitative evaluation of the fleet of boats, and typical ability learning. With the wide application of unmanned boat technology in many fields, the unmanned boat technology has also developed accordingly. The unmanned boat intelligent learning environment is a technical environment designed to cultivate and improve the autonomous learning and intelligent decision-making capabilities of the unmanned boat system. The following is the background technology related to the unmanned boat intelligent learning environment:

[0128] Machine learning and deep learning:

[0129] Machine learning algorithms: Used to enable unmanned boats to learn and extract patterns from data, such as supervised learning, unsupervised learning, and reinforcement learning.

[0130] Deep learning neural networks: Used to process complex perception tasks, such as image recognition, speech recognition, and natural language processing.

[0131] Data acquisition and sensing technology:

[0132] Sensor networks: Used to collect environmental information, such as water temperature, water quality, water flow, weather, etc., as well as the status information of the unmanned boat itself.

[0133] Data preprocessing: Cleaning, normalizing, and denoising data for use by machine learning algorithms.

[0134] Autonomous navigation and path planning:

[0135] SLAM (Simultaneous Localization and Mapping) technology: Used for unmanned boats to build maps and determine their positions in unknown environments.

[0136] Reinforcement learning: Applied to path planning and navigation control to maximize a certain performance metric, such as efficiency, safety, etc.

[0137] Perception technology:

[0138] Computer vision: Unmanned boats can use cameras to perceive the surrounding environment, detect and track targets, and identify obstacles.

[0139] Underwater Sonar: Conduct distance measurement, obstacle detection, and target tracking in the underwater environment.

[0140] Data Analysis and Decision Making:

[0141] Real-time Data Processing: Process data from sensors and external data for decision making, such as weather forecasts and ocean data.

[0142] Reinforcement Learning Decision Making: Machine learning algorithms are used to make decisions based on the current environment and task conditions, such as path selection, target tracking, and obstacle avoidance.

[0143] Simulation and Emulation Environment:

[0144] Virtual Environment: Use virtual reality (VR) or simulation technology to simulate various ocean environments and task scenarios for learning and training.

[0145] Offline Learning: Unmanned boats can conduct offline learning in the simulation environment to reduce the risk of actual operations.

[0146] Cluster Collaboration Technology:

[0147] Cluster Communication and Collaboration: Enable multiple unmanned boats to work together, share information, and complete complex tasks.

[0148] Distributed Learning: Unmanned boats can share knowledge and experience through a communication network to improve the performance of the entire cluster.

[0149] The goal of the intelligent learning environment for unmanned boats is to cultivate the unmanned boat system with autonomous decision-making and adaptation capabilities, enabling it to cope with complex and dynamic ocean environments and perform various tasks, including ocean surveys, search and rescue operations, environmental monitoring, etc. The development of these technical fields provides important support for unmanned boats to play a greater role in practical applications.

[0150] The present invention realizes repeated testing of the boat group by constructing a customizable standardized test environment, synchronously collects relevant data, and realizes full-dimensional testing and quantitative evaluation of the boat group.

[0151] To achieve the above object, the present invention constructs an intelligent learning environment: an intelligent learning environment including the whole process of testing, analysis, and ability improvement, and the main functions are as follows:

[0152] S1. Construct an intelligent learning environment covering the whole process of testing, analysis, and ability improvement. By constructing a customizable standardized test environment, realize repeated testing of the boat group, synchronously collect relevant data, and realize full-dimensional testing and quantitative evaluation of the boat group. The main functions include test plan scenario editing, data synchronization, data set management, data annotation, surface unmanned boat learning environment management, surface unmanned boat intelligent algorithm learning and training functions, data analysis and evaluation.

[0153] Furthermore, the whole process of S1 intelligent learning environment is as follows:

[0154] In a customizable, standardized, and whole-process intelligent learning environment, based on the test plan scenario editing, it is possible to propose test scenarios for different tests, and command and control the tests based on the scenarios; it supports scenario editing, and can set up test items such as water surface obstacles and confrontation targets in the test area.

[0155] It can realize data synchronization, and can realize data synchronization recording and real-time uploading of boat group data during the simulation process; and has the ability to collect unmanned boat motion data, including navigation information, attitude information, and has the ability to collect unmanned boat platform data, including underlying control data, platform vibration data, etc.

[0156] According to the requirements of the surface unmanned boat intelligent algorithm for training data, the historical data collected by the unmanned boat end is managed in a unified manner. And according to the requirements of the surface unmanned boat system intelligent learning for training data, it has the function of manually annotating the video, image and other data of the surface unmanned boat system, providing labeled training data for intelligent learning training.

[0157] Aiming at the management of the intelligent learning environment of surface unmanned boats, it has the functions of creating, importing, querying and modifying the learning environment of surface unmanned boats, and provides a learning environment for intelligent algorithm learning and training. In view of the characteristics of the intelligent algorithm model of the surface unmanned boat system, based on incremental data or learning scenarios, it has the functions of one-click model learning and training and model evaluation, supports the generation of training reports, evaluation reports, etc., supports the visualization of model effects, and realizes the intelligent learning and training upgrade of the surface unmanned boat system model.

[0158] Analyze and evaluate data; support the collection and management of test process data; display key information such as the position and speed of unmanned boats; during the operation of multiple unmanned boats, display the distance between boats in the cluster, calculate the time for formation change, analyze the tracks of multiple boats; analyze simulated platform data and load data; analyze data collected by multiple boats. Support the preservation of evaluation results.

[0159] The beneficial effects of this embodiment are as follows:

[0160] The full-process intelligent learning environment of unmanned boats helps make these autonomous systems more adaptable to changing missions and environmental conditions, improve their performance and availability, reduce operating costs, and enhance their application potential in various application fields. This is of great value for tasks such as marine scientific research, resource surveying, search and rescue, and environmental monitoring.

[0161] Intelligent Decision-making and Autonomy Enhancement: Through continuous data collection and analysis, the intelligent learning environment can help the unmanned boat system improve the decision-making process. This includes better predicting environmental changes, identifying potential hazards, adjusting behavioral strategies, and dealing with problems in real time. The system can learn from historical data and real-time observations, gradually improving the accuracy and efficiency of autonomous decision-making.

[0162] Real-time Adaptability: The unmanned boat can continuously adjust its behavior during operation to cope with changes in the environment and tasks. For example, when the marine conditions suddenly deteriorate or unknown obstacles appear, the unmanned boat can immediately adjust its path or take necessary actions to ensure the successful completion of the task.

[0163] Task Diversity and Complexity Handling: The intelligent learning environment enables the unmanned boat system to handle various types of tasks, from simple patrol and data collection to complex search and rescue and scientific research. The system can adapt according to task requirements, providing the best strategies and action plans.

[0164] Resource Optimization: Environmental perception and data analysis help the unmanned boat system better manage resources, including energy, sensors, and communication bandwidth. This can reduce operating costs, extend mission execution time, and improve the sustainability of the system.

[0165] Training and Simulation: The intelligent learning environment allows for large-scale training and testing in a virtual environment to improve the performance of the unmanned boat without actual sea operations. This reduces risks, cuts the costs of experiments and training, and speeds up the development and verification of new algorithms and strategies.

[0166] Improved Safety and Reliability: The autonomous learning environment helps improve the safety and reliability of the unmanned boat. It can help the system better identify potential safety risks, such as collisions with other vessels or adverse weather conditions. In addition, the system can enhance the reliability of operations by correcting errors or taking emergency measures.

[0167] Cooperative Operations: For a multi-unmanned boat cluster, the intelligent learning environment can assist the unmanned boats in the cluster to work together to complete complex tasks. This cooperative operation ability is of great significance in fields such as search and rescue, environmental monitoring, and military applications.

[0168] Continuous Improvement and Upgrade: The system can continuously learn and improve to adapt to evolving task requirements and technological developments. This enables the unmanned boat system to continuously improve its performance over time and maintain its competitiveness.

[0169] In a specific embodiment, the invention realizes repeated testing of the boat group by constructing a customizable standardized test environment, synchronously collecting relevant data, and achieving full-dimensional testing and quantitative evaluation of the boat group. The environment includes:

[0170] S1 proposes to build an intelligent learning environment: The specific implementation steps of the intelligent learning environment covering the whole process of testing, analysis, and ability improvement are as follows:

[0171] Scenario planning for the experiment:

[0172] Supporting scenario planning for the experiment is one of the key functions in the unmanned boat testing environment, which allows researchers to customize and adjust the experiment scenarios, including surface obstacles and adversarial targets. The following is a detailed description of how to implement support for scenario planning:

[0173] Scenario planning interface: Create a user-friendly scenario planning interface that enables researchers to easily define experiment scenarios. This interface can be a graphical user interface (GUI) or a command-line interface, depending on user needs and system design.

[0174] Adding surface obstacles:

[0175] Allow users to add surface obstacles in the experiment area, such as buoys, islands, vessels, or other obstacles.

[0176] Users can specify the type, location, size, shape, and motion characteristics of the obstacles.

[0177] These obstacles can simulate the obstacles in the actual ocean environment to test the obstacle avoidance ability of the unmanned boat.

[0178] Adding adversarial targets: Users should be able to add adversarial targets, such as simulated hostile unmanned boats or target vessels. Users can define the behavior, speed, heading, and tasks of the adversarial targets. This helps to test the response ability of the unmanned boat system when facing potential threats.

[0179] Setting environmental conditions: Allow users to define environmental conditions, such as wind speed, wave height, water temperature, visibility, etc. These conditions can change dynamically during the experiment to simulate different ocean meteorological conditions.

[0180] Path planning: Provide path planning tools that enable users to specify the starting point, target point, and task route of the unmanned boat. Users can create multiple tasks and paths in the experiment scenario to test the navigation and path planning ability of the unmanned boat under different scenarios.

[0181] Interactive editing and real-time preview: Enable users to edit and preview the experiment scenario in real time for timely adjustment and optimization of the scenario. This can help users better meet the experiment objectives.

[0182] Data recording and playback: Record all parameters and dynamic information during the experiment for subsequent data analysis and evaluation. Users should also be able to playback the experiment to view the behavior and performance of the unmanned boat under different scenarios.

[0183] Documentation and Sharing: Finally, allow users to save and share the edited trial scenarios for sharing with other team members or researchers. This helps with collaboration and knowledge sharing.

[0184] Data Synchronization:

[0185] Data synchronization plays a crucial role in the swarm testing during the simulation process. It can ensure the coordinated work among individual unmanned boats and record and upload relevant data in real time. The following details how to achieve data synchronization and real-time upload:

[0186] Data Synchronization Mechanism:

[0187] Time Synchronization: To ensure that all unmanned boats perform operations at the same time point during the simulation process, a time synchronization mechanism needs to be implemented. This can be achieved through the Network Time Protocol (NTP) or similar time synchronization protocols to ensure that all unmanned boats remain synchronized during the simulation.

[0188] Status Data Synchronization: In addition to time synchronization, it is also necessary to ensure the synchronization of the status data of individual unmanned boats. This includes position, speed, heading, sensor data, etc. Status data synchronization can be achieved through network communication protocols, such as message passing or data sharing.

[0189] Command and Control Synchronization: The coordinated work among unmanned boats requires ensuring the synchronous transmission of command and control information. This can be achieved through a centralized control system or distributed control algorithms to ensure that each unmanned boat performs operations based on the same decisions and tasks.

[0190] Data Recording and Storage:

[0191] Data Recording: During the simulation process, each unmanned boat should record its status data, sensor data, executed tasks, and received commands. These data can be locally recorded on each unmanned boat for subsequent analysis and evaluation.

[0192] Centralized Storage: To achieve real-time data upload, a centralized data storage server or cloud platform can be set up to receive, store, and manage the data of all unmanned boats. This server can run outside or inside the simulation environment.

[0193] Data Analysis and Visualization: Set up data analysis tools on the server side to perform real-time or offline analysis on the uploaded data. These tools can be used to evaluate the performance of unmanned boats, detect problems, or make automated decisions. To monitor the simulation process, a real-time visualization interface can be set up to track the movement, status, and task execution of unmanned boats. This can help operators and researchers understand the behavior of the boat swarm in real time.

[0194] Autonomous cruising: The soft robot uses its buoyancy and soft characteristics to perform autonomous cruising in water. The autonomous control system helps it avoid obstacles.

[0195] Biological image acquisition: The carried camera takes images of marine organisms and records the appearance of the marine ecosystem. The water quality sensor collects data on the water body quality.

[0196] Data transmission: The soft robot transmits the collected biological images and water quality data back to the unmanned boat through the communication system. The unmanned boat transmits the data to the main station for analysis and processing.

[0197] Ability to collect unmanned boat motion data:

[0198] Navigation information data: This data includes the position coordinates (longitude, latitude), speed, acceleration, heading, turning rate, etc. of the unmanned boat. Navigation information data is crucial for understanding the navigation performance and path planning of the unmanned boat. This data can be collected by the GPS and inertial navigation system on the unmanned boat.

[0199] Attitude information data: Attitude information data includes the pitch, roll and yaw attitude angles of the unmanned boat, as well as the related angular velocity data. This data is very important for understanding the stability and control performance of the unmanned boat and is usually collected by an inertial measurement unit (IMU) or gyroscope.

[0200] Sensor data: Various sensor data on the unmanned boat should also be collected, such as underwater sonar data, camera images, water temperature, water quality, meteorological conditions, etc. This data is crucial for environmental perception and task execution.

[0201] Ability to collect platform data:

[0202] Underlying control data: Underlying control data includes the engine status, rudder position, thruster power, battery status, etc. of the unmanned boat. This data helps monitor the performance of the mechanical system and power system of the unmanned boat, as well as the operation of the control system.

[0203] Platform vibration data: Platform vibration data records the vibration and shock conditions of the unmanned boat under different ocean conditions. This is very important for evaluating the stability and durability of the unmanned boat. Usually, an accelerometer or vibration sensor is used to collect this data.

[0204] As Figure 11 shown, Dataset management:

[0205] According to the requirements of the intelligent algorithm of the surface unmanned boat for training data, and for the unified management of the historical data collected at the unmanned boat end, the following key aspects need to be considered:

[0206] Data Diversity: The training data should cover diverse scenarios and environments to ensure that the algorithm performs well under different conditions. This includes different ocean environments, weather conditions, vessel movement patterns, etc. Data diversity helps the algorithm better adapt to various situations in the real world.

[0207] Historical Data Collection: The unmanned boat should be equipped with sensors to collect historical data, including position, speed, attitude, underwater sonar data, sensor readings, etc. This data should be recorded during various tasks and operations.

[0208] As Figure 12 shown, Data Annotation and Comment: Annotating and commenting on the collected historical data is crucial. This means adding metadata to each data point or timestamp, such as event markers, task types, environmental conditions, etc. This helps associate the data with specific contexts and tasks for algorithm training and evaluation. Data annotation is the process of associating samples or data points in the original dataset with relevant information or labels. These labels or comments provide important context information about the data, making the data usable for machine learning and deep learning tasks such as supervised learning, classification, object detection, semantic segmentation, etc. The following is a detailed description of data annotation:

[0209] Types of Labels:

[0210] Classification Labels: These labels are used to classify data into different types or categories. For example, labeling animals in an image as dogs, cats, birds, etc.

[0211] Bounding Box Labels: Used for object detection tasks, marking the location and bounding box of the target in the image. This includes information such as the coordinates, width, and height of the target.

[0212] Semantic Segmentation Labels: These labels are used for pixel-level segmentation tasks, assigning each pixel in the image to different semantic categories. For example, segmenting areas such as roads, vehicles, pedestrians, etc.

[0213] Instance Segmentation Labels: Similar to semantic segmentation, but it differentiates different instances, i.e., separating different objects within the same category.

[0214] Timestamp and Event Labels: For time series data, the labels include timestamps as well as labels related to specific events or states, such as sensor readings, action categories, etc.

[0215] Annotation Tools:

[0216] Manual Annotation: Manual annotation is the most common method, involving manual observation and marking of data. This can be done through graphical interface tools, text editors, or custom software.

[0217] Semi-automatic annotation: Semi-automatic annotation tools combine manual annotation and automatic processing. For example, computer vision algorithms are used to automatically generate initial labels, which are then reviewed and corrected by humans.

[0218] Automatic annotation: For some tasks, machine learning or deep learning models can be used for automatic annotation. This is applicable to large-scale datasets, but high-quality training data is required to build the model.

[0219] Annotation process:

[0220] Data preparation: Before starting annotation, the dataset needs to be prepared, including raw data, images, text, sensor readings, etc.

[0221] Annotation strategy: Define the annotation strategy, including which data needs to be annotated, how to annotate, and whether there are multiple annotators, etc.

[0222] Annotation workflow: For each data point, the annotator annotates according to the annotation strategy and associates the relevant information with the data.

[0223] Quality control: Implement quality control measures, such as double annotation, annotation consistency check, and error repair, to ensure the accuracy of annotation.

[0224] Data format: The annotated data is usually stored in a standard format for subsequent training and evaluation of machine learning models.

[0225] Data quality and calibration: The quality and accuracy of the data are crucial. Ensure the calibration of sensors and the stability of the data acquisition system to reduce noise and errors. Inaccurate data may lead to a decline in the performance of the trained model.

[0226] Data storage and management: All collected historical data should be effectively stored and managed. This can be achieved using a database system or a cloud storage solution to ensure easy access, retrieval, and backup of the data.

[0227] Data privacy and security: Considering the possibility of sensitive data, it is crucial to ensure the privacy and security of historical data. Take appropriate data encryption and access control measures to protect the data from unauthorized access and leakage.

[0228] Continuous update and maintenance: Data management needs to be a continuous process. Over time, new historical data should be continuously collected, and the dataset should be updated and maintained regularly to ensure that the model remains consistent with the latest situation.

[0229] In summary, according to the requirements of the intelligent algorithm for unmanned surface vessels, effectively managing historical data is a key factor in ensuring the performance and reliability of the algorithm. This requires multi-faceted considerations, including data diversity, quality, annotation, privacy protection, and continuous update. Only through carefully managed data can the algorithm perform excellently in practical applications.

[0230] Aiming at the management of the intelligent learning environment of unmanned surface vessels, it has functions such as creating, importing, querying, and modifying the learning environment of unmanned surface vessels, providing a learning environment for the learning and training of intelligent algorithms.

[0231] I. Management of the learning environment of unmanned surface vessels:

[0232] Such an intelligent learning environment management system for unmanned surface vessels will provide a powerful tool for researchers and developers to customize, explore, and optimize the learning and training environment of unmanned surface vessel intelligent algorithms. This not only helps improve the performance of the unmanned vessel system but can also be used for training and education, as well as promoting the development of unmanned surface vessel technology.

[0233] Creating the learning environment:

[0234] Scene customization: Allows users to select or customize different water surface environment scenes, including oceans, lakes, rivers, etc., as well as different weather conditions and environmental factors.

[0235] Geographical feature setting: Allows users to add and edit geographical features such as coastlines, ports, buoys, islands, etc. These features will be used to simulate different water area situations.

[0236] Obstacle and target setting: Users can add water surface obstacles, adversarial targets, or simulate other vessels to test the obstacle avoidance and cooperation capabilities of the unmanned vessel.

[0237] Environmental parameter configuration: Allows users to set environmental parameters such as wave height, wind speed, tide, etc. to simulate different meteorological and ocean conditions.

[0238] Importing the learning environment:

[0239] Data import: Allows users to import map data, weather data, bathymetric data, etc. of the real-world water area from external data sources to more realistically simulate the learning environment.

[0240] Model import: Supports importing existing unmanned surface vessel models, environment models, physical engines, or simulation tools for use in the learning environment.

[0241] Querying the learning environment:

[0242] Environment search: Provides a powerful search function that allows users to query existing learning environments according to different criteria, such as geographical location, scene type, difficulty level, etc.

[0243] Environmental Browsing: Provide a visual environmental browsing interface that allows users to quickly browse and select learning environments.

[0244] History Record: Record the user's query history so that they can easily retrieve previously used environments.

[0245] Modify Learning Environment:

[0246] Environmental Editing: Allow users to edit and customize within existing learning environments, including adding, deleting, or modifying geographical features, obstacles, targets, etc.

[0247] Parameter Adjustment: Allow users to adjust environmental parameters at any time to change ocean and weather conditions, as well as adjust the difficulty of the learning environment.

[0248] Version Control: Provide environmental version control so that users can save different versions of the learning environment and restore to a specific version when needed.

[0249] II. Intelligent Algorithm Learning and Training Function of Unmanned Surface Vehicle:

[0250] For the characteristics of the intelligent algorithm model of the unmanned surface vehicle system, based on incremental data or learning scenarios, it has the functions of one-key model learning and training and model evaluation, supports functions such as training report generation and evaluation report generation, supports the visualization of model effects, and realizes the intelligent learning and training upgrade of the unmanned surface vehicle system model.

[0251] Algorithm Selection and Configuration:

[0252] As Figure 13 shown, Algorithm Library: Provide a variety of different intelligent algorithm options, including deep learning models, reinforcement learning algorithms, path planning algorithms, etc.

[0253] Parameter Configuration: Allow users to configure the parameters of the algorithm according to their research needs and the characteristics of the learning environment to optimize performance.

[0254] Dataset Preparation and Management:

[0255] Dataset Creation:: Allow users to create datasets for training, including input data (sensor data, environmental data) and labels (desired boat behavior, target location, etc.).

[0256] Dataset Annotation: Provide data annotation tools so that users can associate actual data with expected behaviors or results.

[0257] Dataset Management: Allow users to organize, store, and manage datasets, including data version control and data backup.

[0258] Model Training:

[0259] As Figure 14 shown, the training pipeline: A pipeline or workflow that provides training algorithms, including data preprocessing, model construction, loss function definition, etc.

[0260] Model evaluation: Provides tools for evaluating model performance, including evaluation metrics such as accuracy, precision, recall, F1-score, etc.

[0261] Model adjustment and optimization:

[0262] Hyperparameter optimization: Provides tools for hyperparameter optimization to help users find the best combination of model parameters.

[0263] Model transfer learning: Supports transfer learning techniques, allowing users to utilize pre-trained models for transfer learning to accelerate the training of new tasks.

[0264] Model visualization: A technique that presents the internal structure, parameters, training process, and prediction results of machine learning and deep learning models in a graphical or visual manner. It is very important for understanding models, diagnosing problems, optimizing performance, and sharing models with stakeholders.

[0265] Model structure visualization:

[0266] Network structure diagram: For deep learning models, the structure of the neural network can be visualized, including the connections, weights, and activation functions of each layer. This helps to understand the complexity and hierarchical structure of the model.

[0267] Model architecture: Displays the overall architecture of the model, including the input layer, hidden layers, output layer, and their connections. This helps to view the data flow and information transfer of the model.

[0268] Parameter visualization:

[0269] Weight heatmap: The weights of the model can be visualized through a heatmap to help identify which weights have the greatest impact on the output of the model.

[0270] Activation visualization: The activation values of each layer can be visualized to understand how the model processes input data and to help discover potential problems.

[0271] Training process visualization:

[0272] Training loss curve: The curve of the loss function during the training process can be plotted to evaluate the training progress and convergence of the model.

[0273] Learning rate curve: The change in the learning rate can be monitored to ensure the stability of the model during training.

[0274] Implement the support for collecting and managing the data during the test process; be able to display key information such as the position and speed of the unmanned boat; during the operation of multiple unmanned boats, be able to display the distances between the boats within the cluster, be able to calculate the time for formation transformation, be able to analyze the trajectories of multiple boats; be able to analyze the simulated platform data and payload data; be able to analyze the data collected by multiple boats. Support the saving of evaluation results.

[0275] I. Implement data analysis and evaluation

[0276] As Figure 15 shown, data analysis and evaluation play a crucial role in the surface unmanned boat system, which helps to evaluate system performance, improve algorithms, verify models and optimize operations. The following are the key aspects covering data analysis and evaluation:

[0277] Collection and management of data during the test process:

[0278] Data acquisition system: Deploy a data acquisition system to collect the motion data, environmental data and other relevant information of the boat group through sensors (such as GPS, inertial measurement unit, camera, sonar, etc.).

[0279] Data storage: Store the collected data in real-time or in batches to ensure the integrity and availability of the data. Distributed databases or cloud storage can be used to handle large-scale data.

[0280] Data cleaning and preprocessing: Before data analysis, clean and preprocess the original data, including removing outliers, filling in missing data, timestamp alignment, etc.

[0281] Distances between boats within the cluster and formation transformation time:

[0282] Distance measurement: Use the data from GPS or other positioning systems to calculate the distances between the boats within the cluster.

[0283] Formation transformation time: Analyze the change process of the boat group formation, including formation initialization, formation maintenance and formation switching, etc.

[0284] Formation evaluation: According to the goals and requirements of the formation, evaluate the quality and stability of the formation to determine whether optimization is needed.

[0285] Simulated platform data:

[0286] Simulated platform integration: Integrate the data of the simulated platform into the analysis system to simulate different environments and scenarios.

[0287] Recording of simulation results: Record the data generated by the simulated platform, including the state of the virtual environment, object positions, physical properties, etc.

[0288] Comparison between Simulated and Actual Data: Compare the data generated by the simulation platform with the actual data to verify whether the performance of the model in the simulation environment is consistent with the actual environment.

[0289] Data Analysis and Evaluation Tools:

[0290] Data Visualization: Use visualization tools to draw charts such as the trajectories of each boat, formation changes, and environmental data for easy understanding and analysis.

[0291] Data Statistics and Analysis: Use statistical and data analysis methods such as cluster analysis, time series analysis, regression analysis, etc. to extract key information about the behavior of the boat group.

[0292] Machine Learning and Deep Learning: Perform supervised or unsupervised learning on the data to train models to predict and optimize the behavior and performance of the boat group.

[0293] Evaluation Metrics:

[0294] Distance Metrics: Calculate the distances, speeds, and accelerations between each boat to evaluate the coordination and collision avoidance capabilities.

[0295] Formation Evaluation Metrics: Use metrics such as formation error and alignment to evaluate the quality of the formation.

[0296] Performance Metrics: Include task completion time, resource utilization, energy efficiency, etc., which are used to evaluate the overall performance of the unmanned surface vehicle system.

[0297] Report Generation:

[0298] Experimental Report: Automatically generate an experimental report, including data analysis results, evaluation metrics, charts, and visualizations for easy sharing and reporting.

[0299] Suggestions and Optimization: Provide suggestions for the behavior and performance of the boat group and support the generation of optimization solutions.

[0300] Data analysis and evaluation provide key information for the design, optimization, and decision-making of the unmanned surface vehicle system. By comprehensively utilizing sensor data, simulation platform data, and machine learning techniques, the behavior, performance, and safety of the boat group can be continuously improved. This helps to ensure the reliability and efficiency of the unmanned surface vehicle in different tasks and environments.

[0301] Although this specification contains many specific implementation details, these should not be construed as limiting the scope of any invention or the scope of what is claimed, but rather are primarily used to describe the features of specific embodiments of a particular invention. Certain features described in multiple embodiments in this specification can also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment can also be implemented separately in multiple embodiments or in any suitable sub-combination. Additionally, although features may operate in certain combinations as described above and were even initially claimed as such, one or more features from the claimed combination can in some cases be removed from that combination, and the claimed combination can be directed to a sub-combination or a variation of a sub-combination.

[0302] Similarly, although operations are depicted in the drawings in a particular order, this should not be construed as requiring that the operations be performed in the particular order shown or sequentially, or that all illustrated operations be performed, to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Additionally, the separation of the various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0303] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the acts recited in the claims can be performed in a different order and still achieve the desired result. Additionally, the processes depicted in the drawings are not necessarily in the particular order or sequential order shown to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0304] It should be noted that, in this context, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

Claims

1. A testing method, characterized in that: For at least one unmanned boat, the testing method comprises: Acquiring test environment parameters and simulation scenarios corresponding to the test environment parameters; Collecting motion data of at least one unmanned boat in the simulation scenario through a sensor; determining at least one annotation information associated with the motion data; Storing the corresponding annotation information and the motion data in a database in a preset format; Get the training algorithm model; Determine test status information based on the training algorithm model and the corresponding annotation information and the motion data; An evaluation parameter corresponding to at least one of the unmanned boats is determined according to the test state information, the test environment parameters and the motion data.

2. The testing method according to claim 1, characterized in that: The obtaining of the test environment parameters and the simulation scenarios corresponding to the test environment parameters specifically includes: Responsive to the creation instruction, obtaining test environment parameters; Determine environmental information and weather information based on test environment parameters; The simulation scene is created according to the environmental information and the weather information.

3. The testing method according to claim 1, characterized in that: The obtaining of the test environment parameters and the simulation scenarios corresponding to the test environment parameters specifically includes: determining, in response to the import instruction, test environment parameters associated with the environmental information and the weather information stored in the data source; Determine the simulation scenario according to the test environment parameters; or determining, in response to the import instruction, model information corresponding to the test environment parameter, the model information including environment information and weather information; A simulation scenario is determined according to the model information.

4. The testing method according to claim 1, characterized in that: The obtaining of the test environment parameters and the simulation scenarios corresponding to the test environment parameters specifically includes: determining search information related to the test environment parameter in response to the search instruction; A simulation scene corresponding to the search information is determined in a historical scene database.

5. The testing method according to any one of claims 1 to 4, characterized in that: Before collecting the motion data of at least one unmanned boat in the simulation scene through a sensor, the method further includes: receiving an adjustment instruction corresponding to the simulation scenario; The test environment parameters are adjusted according to the adjustment instructions, and a simulation scenario is established according to the adjusted test environment parameters.

6. The testing method according to claim 1, characterized in that: The collecting of motion data of at least one unmanned boat in the simulation scene by means of a sensor specifically includes: Determining environmental data of the environment in which the unmanned boat is located by at least one sensor disposed on the unmanned boat; Determining navigation data of the unmanned boat by at least one positioning sensor disposed on the unmanned boat, the navigation data including position, speed, acceleration, heading, and turning rate; The attitude data of the unmanned boat is determined by an inertial sensor arranged on the unmanned boat, and the attitude data includes pitch, roll, yaw attitude angle, and angular velocity.

7. The testing method according to claim 1, characterized in that: The storing the corresponding annotation information and the motion data in a database in a preset format specifically includes: Obtaining a preset format for storing data in a database and an encryption algorithm for encrypting data; Preprocessing the associated annotation information and the motion data to generate preprocessed data; Encrypting the preprocessed data by the encryption algorithm to generate data to be stored; The data to be stored is stored in the database.

8. The testing method according to claim 1, characterized in that: The obtaining of the training algorithm model specifically includes: Determine an algorithm library storing a plurality of preset algorithm models; Acquire configuration parameters corresponding to at least one of the preset algorithm models; The training algorithm model is determined according to the configuration parameters.

9. The testing method according to claim 1, characterized in that: The determining of the test state information based on the training algorithm model and the corresponding annotation information and the motion data specifically includes: Based on the training algorithm model, the corresponding annotation information and the motion data are input, and simulation data is output using a visualization tool, where the simulation data includes simulation environment information and object position information.

10. The testing method according to claim 9, characterized in that: Determining, according to the test state information, the test environment parameters and the motion data, an evaluation parameter corresponding to at least one of the unmanned boats specifically includes: When the number of the unmanned boat is one, determining the trajectory information of the unmanned boat according to the simulation data; When there are multiple unmanned boats, boat group information of the multiple unmanned boats is determined according to the simulation data, and the boat group information includes coordination information, formation information and system information.

11. A testing device, characterized in that: For at least one unmanned boat, the testing device comprises: An acquisition module, used to acquire test environment parameters and simulation scenarios corresponding to the test environment parameters; A collection module, used for collecting motion data of at least one unmanned boat in the simulation scene through a sensor; a labeling module, configured to determine at least one labeling information associated with the motion data; A storage module, used for storing the corresponding annotation information and the motion data in a database in a preset format; Model acquisition module, used to obtain the training algorithm model; A test module, used to determine test status information based on the training algorithm model and the corresponding annotation information and motion data; An evaluation module is used to determine an evaluation parameter corresponding to at least one of the unmanned boats according to the test state information, the test environment parameters and the motion data.

12. An electronic device, characterized in that: The method comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the test method according to any one of claims 1 to 10.

13. A readable storage medium, characterized in that: The readable storage medium stores a program or an instruction, and when the program or the instruction is executed by a processor, the steps of the testing method according to any one of claims 1 to 10 are implemented.

14. A chip, characterized in that: The chip includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or an instruction to implement the steps of the test method according to any one of claims 1 to 10.

Images