Ship motion prediction method and device, electronic equipment and readable storage medium

Abstract

The invention discloses a ship motion prediction method and device, electronic equipment and a readable storage medium. The ship motion prediction method comprises the steps of obtaining current motion information and current environment information of a target ship; the current motion information and the current environment information are processed through a ship motion prediction model obtained through pre-training, output future motion information is obtained, the ship motion prediction model is constructed based on a motion sub-model and a prediction sub-model obtained through training, the motion sub-model is used for representing physical motion characteristics of the ship, and the prediction sub-model is used for predicting the physical motion characteristics of the ship. The prediction sub-model adopts a gating circulation unit as a model framework and adopts historical ship motion information as a training sample, and the historical ship motion information comprises various parameters which are respectively collected by various sensors arranged on a ship and are related to representing ship motion; determining a future motion track of the target ship by using the future motion information; and adjusting the planned advancing path of the target ship based on the future movement track. By applying the method, the intelligent level of a ship navigation system can be improved, the ship navigation path is optimized, and the safety of sailing on the sea is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a ship motion prediction method, device, electronic device and computer-readable storage medium. Background Art

[0002] Ship path planning and optimization is one of the important issues in the field of navigation. In the complex and changing marine environment, ships need to plan their paths based on real-time environmental information and ship status to ensure navigation safety and efficiency.

[0003] Traditional marine navigation mainly relies on the ship's compass, GPS and other sensor data as well as static information such as nautical charts. However, this information may sometimes not be accurate enough, especially in complex situations such as bad weather, narrow waterways or areas with dense artificial facilities.

[0004] The main methods for ship path planning and optimization are: first, the ship trajectory prediction method based on the dynamic information of AIS (Automatic Identification System). This method uses the real-time data such as ship position and speed provided by the AIS system, combined with historical trajectory data and certain mathematical models, to predict the future path of the ship through probabilistic inference or machine learning methods. However, the adaptability of this method to complex sea conditions needs to be improved, the full utilization of multi-sensor data needs to be strengthened, and there is still room for further improvement in ship trajectory optimization. The second is the path prediction method based on traditional machine learning models: This method uses traditional machine learning algorithms, such as linear regression, decision trees, etc., to model and predict the historical trajectory data of ships.

[0005] Although these methods can predict the future position of the ship to a certain extent, due to the limited ability of these algorithms to model complex nonlinear relationships, they may not be able to accurately capture the complex patterns and laws of ship motion, resulting in inaccurate or inaccurate prediction results. Summary of the invention

[0006] The purpose of this application is to provide a ship motion prediction method, device, electronic device and computer-readable storage medium.

[0007] To achieve the above objectives, the present application provides a ship motion prediction method in a first aspect, the method comprising: Obtain the current movement information and current environment information of the target ship; The current motion information and the current environment information are processed by using the pre-trained ship motion prediction model to obtain output future motion information; wherein the ship motion prediction model is constructed based on the motion sub-model and the trained prediction sub-model, the motion sub-model is used to characterize the physical motion characteristics of the ship, the prediction sub-model uses a gated recurrent unit as a model framework, and uses historical ship motion information as a training sample, and the historical ship motion information includes various parameters related to characterizing the ship motion collected by various sensors installed on the ship; Determine the future motion trajectory of the target ship using the future motion information; The planned path of the target ship is adjusted based on the future motion trajectory.

[0008] Optionally, the various sensors include at least two of the following: Navigation devices, GPS devices, radars, depth sounders, wind direction and anemometers, compasses, accelerometers, gyroscopes, ocean buoys, automatic identification systems AIS, and weighing instruments; Correspondingly, the historical ship motion information collected by various sensors includes at least two of the following: Vessel's latitude and longitude, mass, speed, acceleration, heading, wind speed, wind direction, tide and current, draft, and vessel angle.

[0009] Optionally, before using the historical ship motion information as training samples to train the prediction sub-model, the following is also included: The outliers of various raw data in historical ship motion information are identified by box plots, and the processed data with outliers removed are obtained; For the part of the processed data with missing data, the Kriging interpolation algorithm based on geostatistics is used to interpolate and supplement the data; All types of data in the supplemented data are normalized to obtain the pre-processed historical ship motion information.

[0010] Optionally, the process of establishing the motion sub-model includes the following steps: Based on the classical physical motion characteristics of ships, construct the motion equations that describe the effects of various factors, including hydrodynamics, aerodynamics, inertia and buoyancy, on the motion of ships; Through multi-degree-of-freedom state equations, the state equations of the ship are constructed under multiple motion states including rolling, pitching, lateral movement, longitudinal movement, pitching and yaw; Environmental information including sea conditions, wind speed, wind direction and tidal currents are used as environmental influencing factors; Based on the equation of motion, state equation and environmental influencing factors, a motion sub-model of the ship affected by environmental information is constructed.

[0011] Optionally, the training process of the prediction sub-model includes: Construct an initial prediction sub-model including an input layer, multiple hidden layers with gated recurrent units as gated recurrent units, and an output layer; wherein the hidden layer uses a nonlinear activation function, the output layer uses a linear activation function, the initial prediction sub-model uses a mean square error as a loss function, and the initial prediction sub-model is compiled using an optimization algorithm based on an adaptive learning rate as an optimizer; The training samples are input into the initial prediction sub-model for iterative training. During the iterative training process, the output error is first calculated through forward propagation, and then the weight is adjusted through back propagation. Batch training is used to input the batches of data into the prediction sub-model in training in different training cycles, and finally a trained prediction sub-model is obtained. In the iterative training process, L2 regularization and early stopping method are used to avoid overfitting.

[0012] Optionally, using the future motion information to determine the future motion trajectory of the target ship includes: The ship motion parameters at multiple future time points contained in the future motion information are used to generate the future motion trajectory of the target ship by fitting them in time sequence.

[0013] Optionally, the planned path of the target vessel is adjusted based on the future motion trajectory, including: The planned path is adjusted using the future motion trajectory to obtain an adjusted trajectory; The multi-objective particle swarm optimization algorithm is used to optimize the trajectory after the primary adjustment to obtain the secondary adjusted trajectory; wherein the multi-objective includes multiple targets that the target ship needs to avoid during the movement, including dangerous waters, maritime obstacles, bad weather, offshore buildings, and other ships.

[0014] To achieve the above-mentioned purpose, the present application provides a ship motion prediction device in a second aspect, the device comprising: A ship current information acquisition unit is configured to acquire current motion information and current environment information of the target ship; A motion prediction unit is configured to process current motion information and current environment information using a pre-trained ship motion prediction model to obtain output future motion information; wherein the ship motion prediction model is constructed based on a motion sub-model and a trained prediction sub-model, the motion sub-model is used to characterize the physical motion characteristics of the ship, the prediction sub-model uses a gated recurrent unit as a model framework and uses historical ship motion information as a training sample, and the historical ship motion information includes various parameters related to characterizing the ship motion collected by various sensors installed on the ship; a future motion trajectory determination unit, configured to determine the future motion trajectory of the target ship using the future motion information; The path adjustment unit is configured to adjust the planned travel path of the target ship based on the future motion trajectory.

[0015] Optionally, the various sensors include at least two of the following: Navigation devices, GPS devices, radars, depth sounders, wind direction and anemometers, compasses, accelerometers, gyroscopes, ocean buoys, automatic identification systems AIS, and weighing instruments; Correspondingly, the historical ship motion information collected by various sensors includes at least two of the following: Vessel's latitude and longitude, mass, speed, acceleration, heading, wind speed, wind direction, tide and current, draft, and vessel angle.

[0016] Optionally, the device further comprises: The abnormality identification unit is configured to identify abnormal values ​​of various raw data in the historical ship motion information by using a box plot method before using the historical ship motion information as a training sample to train the prediction sub-model, and obtain processed data after removing the abnormal values; The interpolation supplement unit is configured to interpolate and supplement the part of the processed data with data loss by using the Kriging interpolation algorithm based on geostatistics to obtain supplemented data; The normalization processing unit is configured to perform normalization processing on various types of data in the supplemented data to obtain pre-processed historical ship motion information.

[0017] Optionally, the apparatus further comprises: a motion sub-model building unit configured to build a motion sub-model, and the motion sub-model is further configured to: Based on the classical physical motion characteristics of ships, construct the motion equations that describe the effects of various factors, including hydrodynamics, aerodynamics, inertia and buoyancy, on the motion of ships; Through multi-degree-of-freedom state equations, the state equations of the ship are constructed under multiple motion states including rolling, pitching, lateral movement, longitudinal movement, pitching and yaw; Environmental information including sea conditions, wind speed, wind direction and tidal currents are used as environmental influencing factors; Based on the equation of motion, state equation and environmental influencing factors, a motion sub-model of the ship affected by environmental information is constructed.

[0018] Optionally, the device further includes: a prediction sub-model training unit configured to train and obtain the prediction sub-model, and the prediction sub-model training unit is further configured to: Construct an initial prediction sub-model including an input layer, multiple hidden layers with gated recurrent units as gated recurrent units, and an output layer; wherein the hidden layer uses a nonlinear activation function, the output layer uses a linear activation function, the initial prediction sub-model uses a mean square error as a loss function, and the initial prediction sub-model is compiled using an optimization algorithm based on an adaptive learning rate as an optimizer; The training samples are input into the initial prediction sub-model for iterative training. During the iterative training process, the output error is first calculated through forward propagation, and then the weight is adjusted through back propagation. Batch training is used to input the batches of data into the prediction sub-model in training in different training cycles, and finally a trained prediction sub-model is obtained. In the iterative training process, L2 regularization and early stopping method are used to avoid overfitting.

[0019] Optionally, the future motion trajectory determination unit is further configured to: The ship motion parameters at multiple future time points contained in the future motion information are used to generate the future motion trajectory of the target ship by fitting them in time sequence.

[0020] Optionally, the path adjustment unit is further configured to: The planned path is adjusted using the future motion trajectory to obtain an adjusted trajectory; The multi-objective particle swarm optimization algorithm is used to optimize the trajectory after the primary adjustment to obtain the secondary adjusted trajectory; wherein the multi-objective includes multiple targets that the target ship needs to avoid during the movement, including dangerous waters, maritime obstacles, bad weather, offshore buildings, and other ships.

[0021] To achieve the above-mentioned purpose, the present application provides an electronic device in a third aspect, the electronic device comprising: Memory for storing computer programs; A processor is used to implement the steps of the ship motion prediction method described in any embodiment of the first aspect when executing a computer program stored in a memory.

[0022] To achieve the above-mentioned objectives, the present application provides in a fourth aspect a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of ship motion prediction as described in any embodiment of the above-mentioned first aspect are implemented.

[0023] Compared with the prior art, the ship motion prediction solution provided in the present application not only uses training samples constructed from multi-sensor data installed on the ship to more accurately characterize the historical ship motion conditions, but also uses a motion sub-model that characterizes the physical motion state of the ship and a gated neural unit that is more suitable for processing complex time series data to construct a prediction sub-model, so that the comprehensively constructed ship motion prediction model can predict future ship motion information based on current ship motion information and environmental information, thereby generating a future motion trajectory, so as to facilitate the correction of the currently planned trajectory, thereby improving the intelligence level of the ship navigation system, and further optimizing the ship's navigation path, ultimately ensuring the safety of maritime navigation and improving driving efficiency.

[0024] The present application also provides a ship motion prediction device, an electronic device and a computer-readable storage medium, which have the above-mentioned beneficial effects and are not described in detail here. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0026] Figure 1 A flow chart of a ship motion prediction method provided in an embodiment of the present application; Figure 2 A flowchart of a method for preprocessing historical ship motion information in a ship motion prediction method provided in an embodiment of the present application; Figure 3 A flow chart of a method for constructing a motion sub-model in a ship motion prediction method provided in an embodiment of the present application; Figure 4 A flow chart of a method for determining a future motion trajectory and adjusting a planned path in a ship motion prediction method provided in an embodiment of the present application; Figure 5 A structural block diagram of a ship motion prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] See also Figure 1 , Figure 1 A flowchart of ship motion prediction provided in an embodiment of the present application includes the following steps: Step 101: Acquire current motion information and current environment information of the target ship; This step is intended to enable an executing entity suitable for executing the ship motion prediction method provided by the present application (such as a local server or cloud server for data processing and analysis) to obtain the current motion information and current environmental information of the target ship, wherein the current motion information may include the current mass, current speed, current acceleration, current heading, current draft, current heading angle, etc. of the ship, and the current environmental information may include: current sea conditions, current wind and wave speed, current tidal currents and waves, etc.

[0029] Step 102: Processing current motion information and current environment information using a pre-trained ship motion prediction model to obtain output future motion information; On the basis of step 101, this step aims to use the pre-trained ship motion prediction model to process the current motion information and the current environment information by the above-mentioned execution subject, and obtain the output future motion information. The ship motion prediction model is constructed based on the motion sub-model and the trained prediction sub-model. The motion sub-model is used to characterize the physical motion characteristics of the ship. The prediction sub-model uses the gated recurrent unit (GRU) as the model framework and uses the historical ship motion information as the training sample for training. The historical ship motion information contains various parameters related to the characterization of the ship motion collected by various sensors installed on the ship.

[0030] It should be noted that the reason why GRU is chosen instead of conventional RNN (Recurrent Neural Network) as the prediction model for ship motion is to solve the gradient vanishing and gradient exploding problems encountered by traditional RNN during training. GRU introduces two main gating mechanisms: Reset Gate: determines how much past information needs to be forgotten. It helps the model select old information to be suppressed; Update Gate: controls how much information of the current unit state needs to be updated, and can retain necessary old information while introducing new input. This gating mechanism enables GRU to handle long-term dependencies more effectively. Compared with traditional RNN, it can capture more important information in sequence data. Therefore, it is more suitable for long-term and complex scenarios such as ship motion, and will be able to show better performance in ship position prediction and path optimization. Compared with LSTM (Long Short-Term Memory), GRU has a simpler structure and fewer parameters. GRU uses only two gates, while LSTM uses three gates (input gate, forget gate, and output gate). This simplification means that the GRU is computationally more efficient and is suitable for smaller datasets or when fast training is required.

[0031] Therefore, GRU is more suitable when you need to process long sequence data, want to improve training efficiency, have limited computing resources but still want to obtain good sequence modeling performance, and need to conduct experiments in a short time to explore model performance.

[0032] That is, GRU is chosen as a variant of RNN in order to take advantage of its efficient gating mechanism and less computational complexity to better handle long-term dependencies and information flow problems in sequence data.

[0033] Specifically, the various sensors include: navigation devices, GPS devices, radars, depth sounders, wind direction and anemometers, compasses, accelerometers, gyroscopes, ocean buoys, automatic identification systems AIS and weighing meters. Correspondingly, the historical ship motion information collected by these sensors may include: ship latitude and longitude, mass, speed, acceleration, heading, wind speed, wind direction, tide currents, draft depth, and ship angle.

[0034] Step 103: Determine the future motion trajectory of the target ship using the future motion information; Based on step 102, this step aims to determine the future motion trajectory of the target ship by the above-mentioned execution subject using the future motion information. That is, the future motion information as the prediction result includes the important parameters of the ship's latitude and longitude, speed and heading at multiple time points in the future, and these parameters can be used to form a complete future trajectory of the target ship.

[0035] Step 104: Adjust the planned travel path of the target ship based on the future motion trajectory.

[0036] On the basis of step 103, this step aims to adjust the planned travel path of the target ship based on the future motion trajectory by the above-mentioned execution subject, that is, dynamically adjust the ship's route according to the combination of the future position of the ship and the existing route planning, so as to achieve the optimization of the travel path.

[0037] Compared with the prior art, the ship motion prediction method provided in the present application not only uses training samples constructed from multi-sensor data installed on the ship to more accurately characterize the historical ship motion conditions, but also uses a motion sub-model that characterizes the physical motion state of the ship and a gated neural unit that is more suitable for processing complex time series data to construct a prediction sub-model, so that the comprehensively constructed ship motion prediction model can predict future ship motion information based on current ship motion information and environmental information, thereby generating a future motion trajectory, so as to facilitate the correction of the currently planned trajectory, thereby improving the intelligence level of the ship navigation system, and further optimizing the ship's navigation path, ultimately ensuring the safety of maritime navigation and improving driving efficiency.

[0038] In order to further make the historical ship motion information used as training samples more effective and remove the misleading parts therein, this embodiment also uses Figure 2 A flowchart of a method for preprocessing historical ship motion information is provided, wherein the process 200 includes the following steps: Step 201: using a box plot to identify outliers in various types of original data in historical ship motion information, and obtaining processed data after removing outliers; This step aims to mark the data points exceeding 93% (for example, other reasonable values ​​may also be selected) as outliers by the above-mentioned execution subject based on statistical methods. Specifically, a box plot method may be used to identify outliers and remove them.

[0039] Step 202: For the part of the processed data that has data missing, the Kriging interpolation algorithm based on geostatistics is used to interpolate and supplement to obtain supplemented data; Based on step 01, this step aims to use the Kriging interpolation method to interpolate the lost data in the case of data loss that may occur in large-scale historical data by the above-mentioned execution subject.

[0040] Among them, the Kriging interpolation algorithm is an interpolation method based on geostatistics, which is widely used in geographic information systems (GIS), mining, climatology and other fields that require spatial data analysis. The Kriging interpolation algorithm estimates the value of unknown points by utilizing the spatial correlation between known data points.

[0041] Step 203: normalizing various types of data in the supplemented data to obtain pre-processed historical ship motion information; Based on step 202, this step aims to ensure that the data are within the same magnitude range by the above-mentioned execution subject using Z-score normalization, which is beneficial to subsequent data analysis and model training.

[0042] That is, through the three preprocessing steps of removing outliers, supplementing missing values, and normalizing provided in this embodiment, the data actually finally obtained can make it easier for the model to learn correct knowledge.

[0043] In order to deepen the understanding of how to construct the motion sub-model, this embodiment also Figure 3 A specific implementation method is given, and the process 300 includes the following steps: Step 301: Based on the classical physical motion characteristics of the ship, construct a motion equation that describes the effects of various factors including hydrodynamics, aerodynamics, inertia and buoyancy on the motion of the ship; The equation of motion is a mathematical expression that describes the change in the state of motion of an object under the influence of external forces. For ships, the equation of motion is usually based on Newton's second law of motion, which states that force equals mass times acceleration (F = ma). In the field of navigation, these equations can be more complex because ships are affected by many forces when moving through water. The main influencing factors include: 1. Water power Water resistance: When a ship moves in water, the resistance of water flow to the hull is called water resistance, which is the main factor affecting the speed and navigation stability of the ship.

[0044] Lift: Certain vessels, such as high-speed craft or hydrofoils, may generate lift at certain speeds, which helps to reduce water drag.

[0045] 2. Aerodynamics Wind resistance: When a ship is sailing on the water, the wind it encounters affects its direction and speed. The direction and strength of the wind as well as the area of ​​the ship will affect air resistance.

[0046] Aerodynamic lift: In some cases, the design of a vessel may also result in the generation of aerodynamic lift, affecting its navigation performance.

[0047] 3. Inertia Ship mass: The mass of a ship (including cargo and structure) affects its response to external forces. According to Newton's second law, a larger mass requires a larger force to achieve the same acceleration.

[0048] Turning and accelerating: When a ship turns or accelerates, its inertia will affect the sensitivity and stability of control.

[0049] 4. Buoyancy Buoyancy principle: According to Archimedes' principle, the buoyancy of a ship in water is equal to the weight of the water it displaces. Buoyancy is closely related to the design, load and water level changes of the ship.

[0050] Stability: Buoyancy also affects the stability of the vessel, affecting how it performs in the face of waves and wind.

[0051] Combining the above factors, the equation of motion is usually expressed as a set of differential equations that describe the motion state of the ship under the action of various external forces. By solving these equations, the motion trajectory of the ship in a given time can be obtained, including information such as position, velocity and acceleration.

[0052] That is, this step will help to: 1) Navigation and control: provide accurate navigation guidance for ships, optimize route selection and speed control; 2) Safety management: Helps identify potential collision risks and unstable conditions, thereby improving navigation safety.

[0053] Step 302: constructing a state equation of the ship in multiple motion states including rolling, pitching, lateral movement, longitudinal movement, pitching and yaw by means of a multi-degree-of-freedom state equation; Degrees of freedom refer to the number of independent changes a system can make during motion. For ships, common degrees of freedom include: Roll: The rotation of the ship around its fore-aft axis; Pitch: The rotation of the ship around its transverse axis; Yaw: The rotation of the ship around its vertical axis; Surge: The linear movement of the ship in the longitudinal (fore and aft) direction. Sway: The linear movement of the ship in the lateral (left-right) direction.

[0054] The state equation describes the dynamic behavior of the system, which is usually composed of the equations of motion and mechanical equations. In ship dynamics, the state equation can capture the response of the ship under various motion states.

[0055] By constructing the state equation, the performance of the ship in different motion states can be analyzed. For example, when a ship sails in waves, its roll, pitch, yaw, etc. will be affected, resulting in complex dynamic behavior.

[0056] Constructing a multi-degree-of-freedom state equation usually includes the following steps: 1) Establishing coordinate system: Select an appropriate coordinate system to describe the motion of the ship; 2) Mechanical modeling: Based on Newton's second law or Lagrange's equation, list the motion equations of each degree of freedom of the ship; 3) Coupling effect: Consider the coupling relationship between the various degrees of freedom of motion. For example, roll may affect pitch motion.

[0057] The following analysis can be performed through the multi-degree-of-freedom state equation: 1) Stability analysis: evaluate the stability of the ship under different sailing conditions; 2) Control system design: providing the basis for the ship’s navigation and control systems to ensure safe navigation under various dynamic conditions; 3) Simulation and Optimization: Use numerical simulation tools to test and optimize ship designs.

[0058] Step 303: using environmental information including sea conditions, wind speed and tidal current as environmental influencing factors; Sea conditions, wind speed, wind direction and tidal currents are all important environmental factors that affect the movement of ships. Sea conditions can include wave height, wave period, etc. Wind speed and wind direction refer to the strength and direction of the wind, while tidal currents refer to the speed and direction of the water flow. These factors will directly affect the navigation stability and efficiency of the ship.

[0059] Step 304: Based on the motion equation, the state equation and the environmental influencing factors, a motion sub-model of the ship affected by the environmental information is constructed.

[0060] By combining the above environmental influences with the equations of motion and state, a specific sub-model is created. This means that the model takes into account how the ship's motion will change under specific sea conditions, wind speed, wind direction and tidal currents.

[0061] Based on the above embodiment, this embodiment provides a training process of a prediction sub-model, including the following steps: First, an initial prediction sub-model is constructed, which includes an input layer, multiple hidden layers with gated recurrent units as gated recurrent units, and an output layer. The hidden layer uses a nonlinear activation function, the output layer uses a linear activation function, the initial prediction sub-model uses the mean square error as the loss function, and the initial prediction sub-model is compiled using an optimization algorithm based on an adaptive learning rate as the optimizer. Then, the training samples are input into the initial prediction sub-model for iterative training. During the iterative training process, the output error is first calculated through forward propagation, and then the weight is adjusted through back propagation. In addition, batch training is used to input the batches of data into the prediction sub-model in training in different training cycles, and finally a trained prediction sub-model is obtained. During the iterative training process, L2 regularization and early stopping method are used to avoid overfitting.

[0062] With respect to step 103 and step 104 in process 100, this embodiment further Figure 4 A specific implementation is provided, wherein process 400 includes the following steps: Step 401: Generate the future motion trajectory of the target ship by fitting the ship motion parameters at multiple future time points included in the future motion information in a time sequence; That is, the prediction results include the important parameters of the ship's longitude and latitude, speed and heading at multiple time points in the future. By fitting the parameter information of multiple time points in chronological order, the complete future motion trajectory can be finally obtained.

[0063] Step 402: using the future motion trajectory to adjust the planned path, and obtain an adjusted trajectory; Step 403: Optimize the first-adjusted trajectory using a multi-objective particle swarm optimization algorithm to obtain a second-adjusted trajectory.

[0064] That is, firstly, according to the future position of the ship predicted by the model, the prediction result is combined with the existing route planning to dynamically adjust the ship route and optimize the driving path. Then, the multi-objective optimization algorithm - Particle Swarm Optimization (PSO) can be used to further optimize the path. The main goals include: 1) Avoid potentially dangerous areas: Combine predicted future positions with real-time environmental data to avoid dangerous areas such as obstacles, bad weather, areas with unsuitable water depths, offshore structures, or narrow waterways.

[0065] 2) Minimize conflicts with other ships: Use AIS, radar and other location data to monitor the location and movement of surrounding ships in real time, avoid conflicts with other ships based on prediction results, and reduce collision risks.

[0066] In order to deepen the understanding of the overall solution, this application also provides a complete specific implementation plan through the following examples: 1. Data Collection and Preprocessing 1) Data Collection Collect data from ship sensors and the ship itself, obtain data from various ship sensors on the ship, and integrate navigation information, GPS information, radar information, depth sounder, wind direction and anemometer, compass data, accelerometer, gyroscope, ocean buoy, AIS information and the quality information of the ship itself.

[0067] 2) Feature Selection: The historical data voyage recorder and database collect the ship's historical trajectory data, including the ship's longitude and latitude position, mass, speed, acceleration, heading, heading, wind speed, wind direction, sea waves and currents, draft, and ship angle.

[0068] 3) Remove data outliers, interpolate and normalize data: Based on statistical detection, data points exceeding 93% are marked as outliers. Based on the box plot method, outliers are identified and removed; For large-scale historical data, data loss may occur. The Kriging interpolation method is used to interpolate the missing data.

[0069] Use Z-score normalization to ensure that the data is within the same magnitude range, which is beneficial for subsequent data analysis and model training.

[0070] 2. Ship Motion Modeling 1) Establishment of the equation of motion: Based on the classic ship motion equation, it describes the influence of hydrodynamics, aerodynamics, inertia and buoyancy on the movement of the ship.

[0071] 2) Establishment of state equation: The motion model of the ship is constructed through the six-degree-of-freedom equation, taking into account the motion states such as roll, pitch, lateral, longitudinal, pitch and yaw, and the force conditions are solved according to the characteristics of the ship and the external environmental conditions.

[0072] 3) Introduction of environmental impact factors: Integrate environmental information such as sea conditions, wind speed, wind direction and current to further correct and optimize the ship motion model to improve the accuracy of modeling.

[0073] 3. Training set and test set division The constructed dataset is divided into a training set and a test set, usually with 80% for training and 20% for testing, to ensure the effectiveness of model training and verification.

[0074] 4. Model Building The data set is constructed as in step 1. 1) Model structure selection: Based on the time series characteristics of ship motion data, GRU is selected as the RNN variant. Not only is the GRU structure relatively simple, but it is also computationally efficient and can handle long-term dependencies. Unlike LSTM, GRU controls the flow of information through two gate mechanisms (update gate and reset gate), which reduces computational complexity and shows good performance in ship position prediction and path optimization.

[0075] 2) Number of network layers and nodes: A multi-layer structure is adopted, including an input layer, 3 hidden layers (GRU) and an output layer. The number of hidden layers and the number of nodes in each layer are designed and adjusted according to the data scale and model accuracy requirements.

[0076] 3) Activation function and loss function: Activation function: ReLU (Linear rectification function) nonlinear activation function is used in the hidden layer to enhance the nonlinear expression ability of the model.

[0077] Output layer: Use a linear activation function to output continuous numerical prediction results.

[0078] Loss function: Mean-Square Error (MSE) is used as the loss function to evaluate the deviation between the model prediction value and the actual value to ensure that the model performance meets the accuracy requirements.

[0079] 5. Model Training 1) Model compilation and optimizer selection: When compiling the model, select the optimizer RMSprop (Root Mean Square propagation) to improve training efficiency. The optimizer is responsible for updating the weights of the network and reducing the value of the loss function.

[0080] Set hyperparameters such as learning rate to adjust the training pace of the model to prevent convergence problems caused by training too slowly or too quickly.

[0081] 2) Training process: Input training data for iterative training, calculate the output error through forward propagation, and then adjust the weight through back propagation.

[0082] Using batch training (mini-batch), data is input into the model in batches in each training cycle (epoch) to improve the efficiency and stability of training.

[0083] 3) Overfitting prevention and control: The L2 regularization method is used to randomly ignore the connections of some neurons to prevent the model from being overly dependent on the training data.

[0084] Use the Early Stopping method to stop training in time according to the error change of the validation set to avoid overfitting problems.

[0085] 6. Ship Path Planning and Optimization 1) Predict the future position of the ship: The trained model can predict the future motion state and position of the ship based on input data such as the current ship position, mass, speed, acceleration, heading, wind speed and direction, draft, angle, etc.

[0086] The prediction results include important parameters such as the ship’s longitude and latitude, speed and heading at multiple time points in the future, forming a complete future trajectory.

[0087] 2) Path planning and optimization: Based on the future position of the ship predicted by the model, the prediction result is combined with the existing route planning to dynamically adjust the ship route and optimize the driving path.

[0088] The path is further optimized using the multi-objective optimization algorithm - particle swarm optimization algorithm. The main objectives include: 1) Avoid potentially dangerous areas: Combine predicted future positions with real-time environmental data to avoid dangerous areas such as obstacles, bad weather, areas with unsuitable water depths, offshore structures, or narrow waterways.

[0089] 2) Minimize conflicts with other ships: Use AIS, radar and other location data to monitor the location and movement of surrounding ships in real time, avoid conflicts with other ships based on prediction results, and reduce collision risks.

[0090] 3) Achieve safe and efficient navigation: Safety: Ensure the safety of the ship during navigation and reduce the possibility of accidents by avoiding dangerous areas and conflicts with other ships.

[0091] Efficiency: Dynamically optimize the navigation path according to the real-time environment and ship status, reduce unnecessary detours, improve navigation efficiency, reduce fuel consumption, and achieve voyage optimization.

[0092] That is, the solution provided in this embodiment has advantages over the existing solution in the following aspects: 1. Multi-sensor data: Traditional methods mainly rely on single sensor data or static chart information, which cannot fully reflect complex sea conditions. This solution can provide more comprehensive and real-time environmental perception capabilities and improve the robustness and adaptability of the system by normalizing multiple sensor data (such as integrated navigation information, GPS information, radar information, depth sounder, wind direction and anemometer, compass data, accelerometer, gyroscope, ocean buoy, AIS information and the quality information of the ship itself).

[0093] 2. Use the RNN variant GRU model to process complex time series data: GRU processes sequence data and can better capture the dynamic changes and complex nonlinear relationships of ship motion. Compared with traditional machine learning models, GRU can learn deeper features from historical trajectories and sensor data to improve prediction accuracy.

[0094] 3. Real-time route optimization: The real-time prediction based on the RNN model combined with the optimization algorithm enables ships to dynamically adjust their routes to adapt to changing sea conditions and other external conditions. This adaptive capability significantly improves the safety and efficiency of navigation.

[0095] 4. Comprehensive consideration of multiple environmental influences: The modeling of environmental factors such as sea conditions, tides, wind speed, and wind direction is introduced, so that path planning not only takes into account the ship's own status, but can also be adjusted according to real-time changes in the external environment, avoiding dangerous areas and improving the reliability of navigation decisions.

[0096] 5. Balance between safety and efficiency: This solution not only focuses on improving the accuracy of path prediction, but also realizes dynamic adjustment of the path through optimization algorithms, thereby improving navigation efficiency while ensuring safety, reducing navigation time and fuel consumption.

[0097] 6. Complete process from data preprocessing to model training: The solution covers the steps of data collection, cleaning, normalization, etc. in detail, laying a solid foundation for subsequent model training. The integrity and consistency of the data are ensured by using the Kriging interpolation method.

[0098] 7. Utilize the deep learning capability of the RNN model: Compared with traditional methods, RNN can effectively capture the dynamic characteristics of ship motion. The technical route has detailed designs for RNN model training, verification, and adaptive adjustment to ensure that the model can continuously learn and adapt to new data.

[0099] 8. Combination of path optimization and risk avoidance: Path optimization based on prediction not only improves navigation efficiency but also enhances the safety of ships, which is especially important for ship navigation in complex waterways or severe sea conditions.

[0100] Because the situation is complicated, it is impossible to list them one by one for explanation. Those skilled in the art should be aware that there may be many examples based on the basic method principles provided by this application combined with actual conditions, and all of them should be within the scope of protection of this application without sufficient creative work.

[0101] See below Figure 5 , Figure 5 This is a structural block diagram of a ship motion prediction device 500 provided in an embodiment of the present application. This embodiment exists as a device embodiment corresponding to the above method embodiment. The ship motion prediction device 500 may include: The ship current information acquisition unit 501 is configured to acquire the current motion information and current environment information of the target ship; The motion prediction unit 502 is configured to process the current motion information and the current environment information using a pre-trained ship motion prediction model to obtain output future motion information; wherein the ship motion prediction model is constructed based on a motion sub-model and a trained prediction sub-model, the motion sub-model is used to characterize the physical motion characteristics of the ship, the prediction sub-model uses a gated recurrent unit as a model framework, and uses historical ship motion information as a training sample, and the historical ship motion information includes various parameters related to characterizing the ship motion that are collected by various sensors installed on the ship; A future motion trajectory determining unit 503 is configured to determine the future motion trajectory of the target ship using the future motion information; The path adjustment unit 504 is configured to adjust the planned travel path of the target ship based on the future motion trajectory.

[0102] In some other optional implementations of this embodiment, the various sensors include at least two of the following: Navigation devices, GPS devices, radars, depth sounders, wind direction and anemometers, compasses, accelerometers, gyroscopes, ocean buoys, automatic identification systems AIS, and weighing instruments; Correspondingly, the historical ship motion information collected by various sensors includes at least two of the following: Vessel's latitude and longitude, mass, speed, acceleration, heading, wind speed, wind direction, tide and current, draft, and vessel angle.

[0103] In some other optional implementations of this embodiment, the ship motion prediction device 500 may further include: The abnormality identification unit is configured to identify abnormal values ​​of various raw data in the historical ship motion information by using a box plot method before using the historical ship motion information as a training sample to train the prediction sub-model, and obtain processed data after removing the abnormal values; The interpolation supplement unit is configured to interpolate and supplement the part of the processed data with data loss by using the Kriging interpolation algorithm based on geostatistics to obtain supplemented data; The normalization processing unit is configured to perform normalization processing on various types of data in the supplemented data to obtain pre-processed historical ship motion information.

[0104] In some other optional implementations of this embodiment, the ship motion prediction device 500 may further include: a motion sub-model building unit configured to build a motion sub-model, and the motion sub-model is further configured to: Based on the classical physical motion characteristics of ships, construct the motion equations that describe the effects of various factors, including hydrodynamics, aerodynamics, inertia and buoyancy, on the motion of ships; Through multi-degree-of-freedom state equations, the state equations of the ship are constructed under multiple motion states including rolling, pitching, lateral movement, longitudinal movement, pitching and yaw; Environmental information including sea conditions, wind speed, wind direction and tidal currents are used as environmental influencing factors; Based on the equation of motion, state equation and environmental influencing factors, a motion sub-model of the ship affected by environmental information is constructed.

[0105] In some other optional implementations of this embodiment, the ship motion prediction device 500 may further include: a prediction sub-model training unit configured to train and obtain a prediction sub-model, and the prediction sub-model training unit is further configured to: Construct an initial prediction sub-model including an input layer, multiple hidden layers with gated recurrent units as gated recurrent units, and an output layer; wherein the hidden layer uses a nonlinear activation function, the output layer uses a linear activation function, the initial prediction sub-model uses a mean square error as a loss function, and the initial prediction sub-model is compiled using an optimization algorithm based on an adaptive learning rate as an optimizer; The training samples are input into the initial prediction sub-model for iterative training. During the iterative training process, the output error is first calculated through forward propagation, and then the weight is adjusted through back propagation. Batch training is used to input the batches of data into the prediction sub-model in training in different training cycles, and finally a trained prediction sub-model is obtained. In the iterative training process, L2 regularization and early stopping method are used to avoid overfitting.

[0106] In some other optional implementations of this embodiment, the future motion trajectory determination unit 503 may be further configured to: The ship motion parameters at multiple future time points contained in the future motion information are used to generate the future motion trajectory of the target ship by fitting them in time sequence.

[0107] In some other optional implementations of this embodiment, the path adjustment unit 504 may be further configured to: The planned path is adjusted using the future motion trajectory to obtain an adjusted trajectory; The multi-objective particle swarm optimization algorithm is used to optimize the trajectory after the primary adjustment to obtain the secondary adjusted trajectory; wherein the multi-objective includes multiple targets that the target ship needs to avoid during the movement, including dangerous waters, maritime obstacles, bad weather, offshore buildings, and other ships.

[0108] This embodiment exists as a device embodiment corresponding to the above method embodiment. Compared with the prior art, the ship motion prediction device provided in this embodiment not only uses the training samples constructed by the multi-sensor data set on the ship to more accurately characterize the historical ship motion situation, but also uses the motion sub-model that characterizes the physical motion state of the ship and the gated neural unit that is more suitable for processing complex time series data to construct the prediction sub-model, so that the ship motion prediction model constructed comprehensively can predict the future ship motion information based on the current ship motion information and environmental information, thereby generating the future motion trajectory, so as to correct the current planned trajectory, improve the intelligence level of the ship navigation system, and then optimize the ship's navigation path, and finally ensure the safety of sea navigation and improve the driving efficiency.

[0109] Based on the above embodiments, the present application further provides an electronic device, which may include a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps provided in the above embodiments may be implemented. Of course, the electronic device may also include various necessary network interfaces, power supplies, and other components.

[0110] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by an execution terminal or a processor, the steps provided in the above embodiment can be implemented. The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0111] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0112] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0113] Specific examples are used herein to illustrate the principles and implementation methods of the present application, and the description of the above embodiments is only used to help understand the method and core ideas of the present application. For ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

[0114] It should also be noted that, in this specification, relational terms such as first and second, etc. 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 terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or equipment including the elements.

Claims

1. A ship motion prediction method, characterized in that: include: Obtain the current movement information and current environment information of the target ship; The current motion information and the current environment information are processed by using a pre-trained ship motion prediction model to obtain output future motion information; wherein the ship motion prediction model is constructed based on a motion sub-model and a trained prediction sub-model, the motion sub-model is used to characterize the physical motion characteristics of the ship, the prediction sub-model uses a gated recurrent unit as a model framework, and uses historical ship motion information as a training sample, the historical ship motion information includes various parameters related to characterizing the ship motion collected by various sensors installed on the ship; Determining a future motion trajectory of the target ship using the future motion information; The planned travel path of the target ship is adjusted based on the future motion trajectory.

2. The method according to claim 1, characterized in that: Each of the sensors includes at least two of the following: Navigation devices, GPS devices, radars, depth sounders, wind direction and anemometers, compasses, accelerometers, gyroscopes, ocean buoys, automatic identification systems AIS, and weighing instruments; Correspondingly, the historical ship motion information collected by various sensors includes at least two of the following: Vessel's latitude and longitude, mass, speed, acceleration, heading, wind speed, wind direction, tide and current, draft, and vessel angle.

3. The method according to claim 2, characterized in that Before using the historical ship motion information as the training sample to train the prediction sub-model, the method further includes: Using a box plot method to identify outliers in various types of original data in the historical ship motion information, and obtaining processed data after removing the outliers; For the part of the processed data where there is data loss, the Kriging interpolation algorithm based on geostatistics is used to interpolate and supplement to obtain supplemented data; The various types of data in the supplemented data are normalized to obtain pre-processed historical ship motion information.

4. The method according to claim 1, characterized in that: The process of establishing the motion sub-model comprises the following steps: Based on the classical physical motion characteristics of ships, construct the motion equations that describe the effects of various factors, including hydrodynamics, aerodynamics, inertia and buoyancy, on the motion of ships; Through multi-degree-of-freedom state equations, the state equations of the ship are constructed under multiple motion states including rolling, pitching, lateral movement, longitudinal movement, pitching and yaw; Environmental information including sea conditions, wind speed, wind direction and tidal currents are used as environmental influencing factors; Based on the motion equation, the state equation and the environmental influencing factors, a motion sub-model of the ship affected by the environmental information is constructed.

5. The method according to claim 1, characterized in that The training process of the prediction sub-model includes: Constructing an initial prediction sub-model including an input layer, a plurality of hidden layers served by the gated recurrent units, and an output layer; wherein the hidden layer uses a nonlinear activation function, the output layer uses a linear activation function, the initial prediction sub-model uses a mean square error as a loss function, and the initial prediction sub-model is compiled using an optimization algorithm based on an adaptive learning rate as an optimizer; The training samples are input into the initial prediction sub-model for iterative training, and in the iterative training process, the output error is first calculated by forward propagation, and then the weight is adjusted by back propagation, and batch training is adopted to input each batch of data into the prediction sub-model in training in different training cycles, and finally a trained prediction sub-model is obtained; wherein, in the iterative training process, L2 regularization and early stopping method are adopted to avoid overfitting.

6. The method according to any one of claims 1 to 5, characterized in that: The determining the future motion trajectory of the target ship by using the future motion information includes: The ship motion parameters at multiple future time points included in the future motion information are used to generate the future motion trajectory of the target ship by fitting them in time sequence.

7. The method according to any one of claims 1 to 5, characterized in that: The adjusting the planned travel path of the target ship based on the future motion trajectory includes: Using the future motion trajectory to adjust the planned travel path, to obtain an adjusted trajectory; The first adjusted trajectory is optimized using a multi-objective particle swarm optimization algorithm to obtain a second adjusted trajectory; wherein the multi-objectives include multiple targets that the target ship needs to avoid during its travel, including dangerous waters, marine obstacles, severe weather, offshore buildings, and other ships.

8. A ship motion prediction device, characterized in that: include: A ship current information acquisition unit is configured to acquire current motion information and current environment information of the target ship; A motion prediction unit is configured to process the current motion information and the current environment information using a pre-trained ship motion prediction model to obtain output future motion information; wherein the ship motion prediction model is constructed based on a motion sub-model and a trained prediction sub-model, the motion sub-model is used to characterize the physical motion characteristics of the ship, the prediction sub-model uses a gated recurrent unit as a model framework and uses historical ship motion information as a training sample, the historical ship motion information includes various parameters related to characterizing the ship motion collected by various sensors installed on the ship; a future motion trajectory determining unit, configured to determine the future motion trajectory of the target ship using the future motion information; A path adjustment unit is configured to adjust the planned travel path of the target ship based on the future motion trajectory.

9. An electronic device, characterized in that: include: Memory, for computer programs; A processor, configured to implement the steps of the ship motion prediction method as described in any one of claims 1 to 7 when executing a computer program stored in the memory.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and after being executed by a processor, the computer program can implement each step of the ship motion prediction method as described in any one of claims 1 to 7.

Images