Radar trial ship operation method and device based on self-adaptive identification of dynamic characteristics of ship

Abstract

The invention discloses a radar trial ship operation method and device based on ship dynamic characteristic self-adaptive identification, and belongs to the technical field of ship control analogue simulation. The method comprises the steps that modeling is conducted on historical ship navigation data through machine learning identification, a ship manipulation motion model is constructed, the ship manipulation motion model is used for predicting motion parameters and tracks of a ship, and the historical ship navigation data comprise the position, the speed and the course of the ship; the motion parameters of the ship comprise the rotation radius, the acceleration and the bow turning rate of the ship; on the basis of the trained ship manipulation motion model, predicted ship motion parameters and tracks are obtained; and obtaining a ship trial operation instruction, and determining the feasibility of the radar trial operation ship scheme according to the collision avoidance safety threshold by combining the predicted ship motion parameters and trajectory. According to the method, the ship manipulation motion model obtained by machine learning identification modeling is mainly utilized, the predicted ship dynamic parameters are used as a simulation basis, and the accuracy of a radar trial ship manipulation collision avoidance scheme is improved.

Description

Technical Field

[0001] The present invention relates to the field of ship control, and in particular to a radar test ship maneuvering method and device based on adaptive identification of ship dynamic characteristics. Background Art

[0002] Navigation radar is one of the main navigation equipment of ships. It plays a vital role in ship observation, positioning, guiding ships to and from ports and navigating in narrow waterways. Its role in ship collision avoidance is irreplaceable. The radar test ship function can help drivers verify the possibility of adopting avoidance plans through graphic simulation. The International Maritime Organization clearly requires in the radar performance standard MSC.192 (79) resolution and the SOLAS Convention that ships with a gross tonnage of more than 10,000 tons must be equipped with radars with a test ship function, and this function should include simulation of the dynamic characteristics of the ship and provide a countdown to the simulation time until the ship maneuvers. However, the existing radar test ship function requires the ship operator to manually input real-time ship parameters such as the ship's bow turning rate and speed change rate. However, the ship's dynamic characteristic parameters change with the steering instructions such as the rudder angle and the car bell command. The ship operator cannot accurately obtain and input the ship's dynamic characteristic parameters on the radar, resulting in the ship operator being unable to use the radar test ship to obtain an accurate and effective collision avoidance plan. Summary of the invention

[0003] According to the technical problem that the uncertainty of artificial input data affects the simulation effect of radar test maneuvering, a radar test maneuvering method and device based on adaptive identification of ship dynamic characteristics is provided. The present invention mainly uses the ship maneuvering motion model obtained by machine learning identification modeling, takes the predicted ship dynamic parameters as the simulation basis, and improves the accuracy of the radar test maneuvering collision avoidance plan.

[0004] The technical means adopted by the present invention are as follows: A radar trial ship maneuvering method based on adaptive identification of ship dynamic characteristics comprises the following steps: The ship's historical navigation data is modeled by machine learning identification, so as to construct a ship maneuvering motion model, wherein the ship maneuvering motion model is used to predict the ship's motion parameters and trajectory; the ship's historical navigation data includes the ship's position, heading, speed, rudder angle, propeller speed and torque, and the ship's motion parameters include the ship's turning radius, acceleration, bow turning rate and lag distance; the ship's maneuvering motion model includes a one-dimensional convolution layer, an LSTM layer, a multi-head attention mechanism, a first fully connected layer and a second fully connected layer, wherein the ship's historical navigation data is processed by the one-dimensional convolution layer to generate input data, and the input data is input into the LSTM layer, and the LSTM layer introduces a Dropout mechanism, and then the output of the LSTM layer is dynamically weighted according to the importance of the ship's maneuvering motion data through the multi-head attention mechanism, and the data processed by the multi-head attention mechanism is input into the first fully connected layer, and the first fully connected layer introduces a Dropout mechanism, and finally the information processed by the first fully connected layer is mapped to the prediction output space through the second fully connected layer; Based on the ship maneuvering motion model after training and optimization, the predicted ship motion parameters and trajectory are obtained, and the predicted ship motion parameters and trajectory are used to simulate the dynamic characteristics of the ship during the radar trial maneuvering process, providing a basis for the verification of the radar trial maneuvering collision avoidance plan; The ship trial maneuvering instruction is obtained, and the feasibility of the radar trial maneuvering plan is determined according to the radar trial maneuvering ship collision avoidance safety threshold in combination with the predicted ship motion parameters and trajectory. The radar trial maneuvering ship collision avoidance safety threshold includes the closest encounter distance between the own ship and the target ship.

[0005] Furthermore, it also includes dynamically displaying the predicted trajectory of the ship and visually showing the execution results of the ship's current collision avoidance plan.

[0006] Furthermore, the training optimization step of the ship maneuvering motion model includes: Preprocess the historical ship navigation data to construct training data sets, verification data sets, and test data sets; Training the ship maneuvering motion model based on a training data set; Performing hyperparameter tuning on the trained ship maneuvering motion model based on the validation data set; The prediction performance of the optimized ship maneuvering motion model is verified based on the test data set.

[0007] Furthermore, the training optimization step of the ship maneuvering motion model also includes: Automatically update and optimize the model based on real-time updated ship data.

[0008] The present invention also discloses a radar test ship maneuvering device based on adaptive identification of ship dynamic characteristics, which is used to implement any of the above methods, including: A ship maneuvering motion modeling module, which uses machine learning to identify historical ship navigation data to model the ship, thereby constructing a ship maneuvering motion model, and is used to output predicted ship motion parameters and trajectories based on the constructed ship maneuvering motion model; the historical ship navigation data includes the ship's position, speed and heading, and the ship's motion parameters include the ship's turning radius, acceleration and bow turning rate; A pre-manipulation control panel module, which is used to obtain pre-manipulation commands and display ship motion parameters predicted by identification modeling; The dynamic generation module of the trial maneuvering plan is used to obtain the ship's trial maneuvering instructions, combine the predicted ship motion parameters and trajectory, and determine the feasibility of the trial maneuvering plan according to the collision avoidance safety threshold. On the other hand, it is used to dynamically display the predicted trajectory of the ship and visualize the execution results of the ship's current collision avoidance plan.

[0009] Further, the trial ship maneuvering plan dynamic generation module includes a trial ship maneuvering function module and a trial ship maneuvering plan display; The trial ship maneuvering function module predicts the future motion data of the ship and calculates the corresponding motion trajectory based on the obtained ship maneuvering motion model and the ship pre-maneuvering command input by the pre-maneuvering control panel; The trial ship maneuvering plan display dynamically displays the ship motion trajectory under different pre-maneuvering control command conditions, so that the trial ship maneuvering plan is accurately presented on the display screen.

[0010] Furthermore, when implementing the radar trial ship maneuvering, the operator generates an input signal through the operation area on the pre-manipulation control panel, and the input signal is captured by the circuit of the pre-manipulation control panel, converted into a digital signal, and then processed through signal amplification, filtering, and analog-to-digital conversion steps, and then transmitted to the trial ship maneuvering function module through the bus. After receiving the signal from the pre-manipulation control panel, the trial ship maneuvering function module determines the operation to be performed according to the preset program and logical judgment, and dynamically displays it through the trial ship maneuvering plan display.

[0011] Compared with the prior art, the present invention has the following advantages: 1. The present invention uses an artificial intelligence algorithm model to design a machine learning model that can automatically predict the motion parameters of a ship. It can accurately simulate the dynamic characteristics of the current ship, accurately predict the future motion parameters and trajectory of the ship according to the input ship pre-manipulation command, verify the feasibility of the current collision avoidance plan, and effectively improve the operability and accuracy of the radar test ship.

[0012] 2. The present invention utilizes electronic engineering technology to develop and design a set of radar trial ship maneuvering pre-manipulation control panels and trial ship maneuvering plan displays, which can more conveniently complete the setting and adjustment of ship motion parameters, and dynamically present the collision avoidance effect under the current pre-manipulation command conditions through the display platform, thereby improving the efficiency of the radar trial ship maneuvering function in providing feasible collision avoidance plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0014] Figure 1 The present invention is a flow chart of a radar trial ship maneuvering method based on adaptive identification of ship dynamic characteristics in an embodiment of the present invention.

[0015] Figure 2 The present invention provides a method for constructing a ship maneuvering motion model in an embodiment of the present invention.

[0016] Figure 3 This is a diagram of the ship maneuvering motion model architecture in an embodiment of the present invention.

[0017] Figure 4 The present invention is a diagram showing the architecture of a radar test ship maneuvering device based on adaptive identification of ship dynamic characteristics in an embodiment of the present invention.

[0018] Figure 5 This is a design effect diagram of the pre-manipulation control panel and display platform in an embodiment of the present invention.

[0019] Figure 6 This is a rendering of the ship motion state simulated by the radar test ship at time T0 in an embodiment of the present invention.

[0020] Figure 7 This is a rendering of the ship motion state simulated by the radar test ship at time T1 in an embodiment of the present invention.

[0021] Figure 8 This is a rendering of the ship motion state simulated by the radar test ship at time T2 in an embodiment of the present invention.

[0022] Fig. 9 This is a rendering of the ship motion state simulated by the radar test maneuvering ship at time T3 in an embodiment of the present invention.

[0023] Fig.10 This is a rendering of the ship motion state simulated by the radar test ship at time T4 in an embodiment of the present invention.

[0024] Fig.11 This is a rendering of the effect when the collision avoidance solution is not feasible in the embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] like Figure 1 As shown, an embodiment of the present invention provides a radar trial ship maneuvering method based on adaptive identification of ship dynamic characteristics, comprising the following steps: S1. Modeling the historical navigation data of the ship using machine learning identification, thereby constructing a ship maneuvering motion model, wherein the ship maneuvering motion model is used to predict the motion parameters and trajectory of the ship. In an embodiment of the present invention, the historical navigation data of the ship includes the ship's position, heading, speed, rudder angle, propeller speed and torque, and the motion parameters of the ship include the ship's turning radius, acceleration and bow turning rate. Furthermore, the motion parameters of the ship may also include the lag distance displayed on the pre-maneuvering control panel, that is, the distance traveled from the time the rudder is manipulated to the time the ship starts to turn.

[0027] Specifically, the ship maneuvering motion model is as follows Figure 3 As shown in the figure, a machine learning model is used, specifically a combined model consisting of a convolutional neural network, a long short-term memory network and a multi-head attention mechanism, to train and establish a prediction model based on real-time data collected from ships, such as ship position, speed, heading and other dynamic information. By learning from historical navigation data, the model can identify and predict information such as ship speed and trajectory changes under different parameter conditions in real time.

[0028] First, the input data is processed through a one-dimensional convolutional layer, which is used to extract important features from the original time series data, such as the local change pattern of the ship's speed, heading, rudder angle and other data at different times, which will provide refined input data for the following LSTM layer. Then, the data processed by the CNN layer is input into the LSTM layer. LSTM is specially designed to process time series data. It can effectively process the time series data in the ship's maneuvering movement and can capture the dependency of the ship's motion state on a longer time scale, such as the long-term changes in the ship's navigation trajectory under complex sea conditions, and the long-term impact of different stages of maneuvering instructions on the subsequent movement of the ship, which makes up for the shortcomings of CNN in processing long-term dependencies. In addition, the Dropout mechanism is introduced in the LSTM layer to enhance the generalization ability of the model, avoid overfitting, and ensure that the model can still perform well when facing new and unseen data. Then, a multi-head attention mechanism is added to the model. Its function is to weight the output of the LSTM layer. It can dynamically assign weights according to the importance of the ship maneuvering motion data, focusing on the most critical information for modeling the ship maneuvering motion, thereby more accurately capturing the key features and important patterns in the ship maneuvering motion. This weighted processing makes the model more accurate in analysis and prediction. Afterwards, the data processed by the attention mechanism is sent to a fully connected layer to aggregate and map the information to an intermediate representation space. Through the application of the ReLU activation function, the model can enhance the ability to capture nonlinear relationships and further enhance the accuracy of the prediction. At the same time, at this stage, the model also introduces the Dropout mechanism to enhance its generalization ability. Finally, another fully connected layer maps the processed information to the prediction output space. The role of this layer is to convert all the learning results of the model into specific prediction values ​​to provide support for the final decision.

[0029] At this point, the ship maneuvering motion model can be obtained, and the corresponding pre-manipulation commands are input into the model. The pre-manipulation instructions include rudder angle and speed, and specific predicted values ​​are obtained. The predicted trajectory can be solved based on the predicted values, as well as the four parameters displayed on the pre-manipulation control panel: distance, ship turning radius, acceleration, and bow turn rate.

[0030] The structure of the convolutional neural network in the combined model mainly includes: 1. One-dimensional convolution layer (Conv1D):

[0031] in: is at the time step Output at is the activation function; is the weight of the convolution kernel; Represents the convolution operation; is at the time step The input at is the size of the convolution kernel; is the bias term.

[0032] The one-dimensional convolution layer uses 128 convolution kernels, the convolution kernel size is 4, the activation function is ReLU, the filling method is "same filling", and the window size of the pooling layer is 2. The time step of the model after pooling is multiplied by 1 / 2, and the number of features and channels remain constant.

[0033] 2. Maximum pooling layer (MaxPooling1D):

[0034] in: is at the time step Output at is at the time step arrive The input between is the size of the pooling window.

[0035] CNN, through its advantages of local feature extraction, dimensionality reduction and feature fusion, and multi-level feature learning, can extract more refined and effective input features from the high-dimensional, time series data of ship manipulation.

[0036] The structure of the long short-term memory network in the combined model mainly includes: 1. Multi-layer long short-term memory network (LSTM): Forget Gate:

[0037] in: is at the time step The output of the forget gate; is the activation function; is the weight matrix of the forget gate; is the bias term of the forget gate; is at the time step The state of forgetfulness; is at the time step The output at .

[0038] Input Gate:

[0039] in: is at the time step The output of the forget gate; is the weight matrix of the forget gate; is the bias term of the forget gate.

[0040] Candidate Cell State:

[0041] in: is at the time step The output of the forget gate; is the hyperbolic tangent activation function; is the weight matrix of the forget gate; is the bias term of the forget gate.

[0042] Cell State Update:

[0043] in: is at the time step The unit status; is at the time step The unit status.

[0044] Output Gate:

[0045] in: is at the time step The output of the forget gate; is the weight matrix of the forget gate; is the bias term of the forget gate.

[0046] Hidden State Update:

[0047] in: is at the time step The hidden state of .

[0048] 2. Dropout: Dropout is an effective regularization method that can prevent overfitting. Dropout randomly discards the output of some neurons, making the neural network independent of specific neurons during training, thereby enhancing the generalization ability of the model. Suppose the input vector before the dropout layer is , the discard probability is , indicating that each neuron has The probability of being discarded (set to 0). The operation of the discard layer can be expressed as:

[0049] in: is the output of the discard layer. The activation value of each neuron; is a binary random variable, 1 means the neuron is retained. 0 means discarding neurons. is calculated based on the drop probability randomly generated, and Each element of is independent of the other elements.

[0050] In this embodiment, LSTM has a three-layer structure. The first layer of LSTM can help capture the local time dependency and short-term pattern of sequence data, learn how to affect the instantaneous movement of the ship according to past maneuvering instructions, and set the number of units to 256; the second layer of LSTM is used to further process the features from the first layer, capture dependencies within a longer time range, and set the number of units to 128; the third layer of LSTM further abstracts and integrates the information extracted by the first two layers of LSTM, learns more complex long-term dependencies, and sets the number of units to 32. In the three layers of LSTM in this model, the Dropout mechanism is introduced only in the first and second layers. These two layers are usually used to extract local features of input data and lower-order temporal dependencies. Adding Dropout can effectively reduce the complexity of the model and prevent overfitting; while the third layer of LSTM integrates the information of the first two layers and performs long-term dependency modeling. Excessive introduction of Dropout may affect its learning of long-term dependencies.

[0051] The movement of the ship is not only affected by the current control instructions, but also by the cumulative effect and time lag effect of past instructions. The three-layer LSTM design enables the model to capture dependencies at different time scales layer by layer.

[0052] The structure of the multi-head attention mechanism in the combined model mainly includes: 1. Linear transformation: The input query, key, and value are first passed through different linear transformation layers to get the projection of each head:

[0053] in: are the original query, keys, and values; It is The weight matrix of each head.

[0054] 2. Independent calculation of attention:

[0055] in: is the dimension of the key vector, express The transpose of .

[0056] 3. Splicing and linear mapping:

[0057] in: It is Output of the head; is a linear layer that projects the merged output back to the original feature space.

[0058] In this application, the importance of ship maneuvering motion data is obtained through the attention mechanism, that is, through the features and weights learned by the model during the training process, the attention mechanism can give different weights to the input data of different time steps, thereby measuring the influence of each time step on the ship's maneuvering motion. Specifically, the dynamic allocation of weights is achieved through a multi-head self-attention mechanism. That is, the feature vector of each input time step will be associated with a query (Query), key (Key), and value (Value), and the attention weight of each time step is calculated. Specifically, the dot product of the query and the key is first calculated, and then normalized to obtain the attention score for each time step. After that, the values ​​are weighted by these attention scores to obtain a weighted feature representation.

[0059] In this application, the ship maneuvering motion model also has online learning and incremental training functions, which can automatically update and optimize the model according to the real-time updated ship data to ensure that it always maintains efficient and accurate prediction capabilities. Figure 2 As shown, the construction and optimization of the ship maneuvering motion model includes the following steps: a. Preprocess the historical ship navigation data to construct training data set, verification data set and test data set. The preprocessing steps include data analysis, feature extraction and data partitioning. Specifically: Data analysis: Not all historical ship navigation data can be used to train the model. The original data may contain noise, missing values ​​or inconsistent data points. Preliminary analysis is needed to identify abnormal data. Filtering and missing data filling methods can be used to ensure data quality.

[0060] Feature extraction: There are many types of feature parameters in the historical navigation data of ships. It is necessary to extract features that are meaningful for modeling and can effectively characterize the ship's maneuvering behavior, namely, ship speed, rudder angle, and heading.

[0061] Data division: The rule is that the first 40% of the data set is the training set, the middle 20% is the validation set, and the last 40% is the test set.

[0062] b. Training the ship maneuvering motion model based on the training data set; c. Performing hyperparameter tuning on the trained ship maneuvering motion model based on the validation data set; d. Verify the prediction performance of the optimized ship maneuvering motion model based on the test data set.

[0063] Furthermore, it also includes: e. Automatically update and optimize the model based on real-time updated ship data.

[0064] Based on the ship maneuvering motion model, the ship's maneuvering motion model can be trained according to the navigation data of each ship. Based on the maneuvering motion model, the trial ship maneuvering function module can more accurately simulate the dynamic characteristics of the ship, provide more precise, timely and effective collision avoidance solutions, and ensure the safety of ship navigation.

[0065] S2. Based on the ship maneuvering motion model after training optimization, the predicted ship motion parameters and trajectory are obtained, and the predicted ship motion parameters and trajectory are used to simulate the dynamic characteristics of the ship during the radar trial maneuvering process, providing a basis for the verification of the radar trial maneuvering ship collision avoidance plan.

[0066] S3. Obtain the ship trial maneuvering instruction, combine the predicted ship motion parameters and trajectory, and determine the feasibility of the trial maneuvering plan according to the trial maneuvering ship collision avoidance safety threshold, wherein the trial maneuvering ship collision avoidance safety threshold includes the closest encounter distance between the own ship and the target ship.

[0067] Furthermore, the method further comprises: S4. Dynamically display the predicted trajectory of the ship and visualize the execution results of the ship's current collision avoidance plan.

[0068] like Figure 4 As shown, another aspect of an embodiment of the present invention further provides a radar trial ship maneuvering device based on adaptive identification of ship dynamic characteristics, which is used to implement the above method, including a ship maneuvering motion modeling module, a pre-manipulation control panel module and a trial ship maneuvering plan dynamic generation module.

[0069] The ship maneuvering motion modeling module is equipped with the above-mentioned ship maneuvering motion model, and outputs predicted ship motion parameters and trajectory based on the ship maneuvering motion model; the ship's historical navigation data includes the ship's position, speed and heading, and the ship's motion parameters include the ship's turning radius, acceleration and bow turning rate.

[0070] The dynamic generation module of trial maneuvering plan is used to obtain the trial maneuvering instructions of the ship, combine the predicted ship motion parameters and trajectory, and determine the feasibility of the trial maneuvering plan according to the collision avoidance safety threshold. On the other hand, it is used to dynamically display the predicted trajectory of the ship and visualize the execution results of the current collision avoidance plan of the ship.

[0071] Specifically, the dynamic generation module of the trial ship maneuvering plan includes a trial ship maneuvering function module and a trial ship maneuvering plan display. The trial ship maneuvering function module predicts the future motion data of the ship and solves the corresponding motion trajectory based on the obtained ship maneuvering motion model and the ship pre-maneuvering command input by the pre-maneuvering control panel. The ship maneuvering plan display is used to dynamically display the ship motion trajectory under different pre-maneuvering control command conditions, so that the trial ship maneuvering plan can be accurately presented on the display screen.

[0072] The pre-manipulation control panel module is used to obtain pre-manipulation commands on the one hand, and to display the ship motion parameters predicted by identification modeling on the other hand.

[0073] Specifically, the pre-manipulation control panel module includes a ship trial steering module, a trial steering clock module, a time delay setting and other ship motion parameter display modules. The trial steering module can simulate the rudder system of a real ship and control the input of the ship's rudder angle parameters to simulate the predicted trajectory of the ship under different rudder angles. The trial steering clock module can simulate the clock system of a real ship and control the increase or decrease of the ship speed to simulate the predicted trajectory of the ship under different ship speeds. The time delay setting can adjust the time delay of the trial maneuvering ship. Other ship motion parameter display modules include parameters such as the ship's turning radius, acceleration, and bow turning rate. These parameters are obtained based on the input pre-manipulation command and then predicted by the ship maneuvering motion model.

[0074] The operator generates input signals by operating the buttons or knobs on the equipment panel. These input signals are captured by the circuit of the control panel and converted into digital signals. After being processed by steps such as signal amplification, filtering, and analog-to-digital conversion, they are transmitted to the trial ship operation function module through the bus. After receiving the signal from the control panel, the module determines the operation to be performed based on the preset program and logical judgment, and dynamically displays it on the trial ship operation plan display.

[0075] The radar test ship operation is carried out based on the above device. The specific steps include: 1. Turn on the trial ship handling device.

[0076] 2. The operator inputs a pre-operation command on the pre-operation control panel to generate an input signal, which is captured by the circuit of the control panel and converted into a digital signal, and then processed through signal amplification, filtering, and analog-to-digital conversion steps.

[0077] 3. The computer on board the test ship operation function module receives the signal from the control panel and calculates the ship's motion parameters and trajectory according to the preset program and logical judgment.

[0078] 4. The trial maneuvering plan display visualizes the collision avoidance effect of the ship under the current pre-maneuvering command.

[0079] 5. If the collision avoidance solution is not feasible, repeat steps 2, 3, and 4 until a feasible collision avoidance solution is achieved.

[0080] like Figure 5 This is the design drawing of the radar trial ship pre-manipulation control panel and the trial ship maneuvering plan display. The specific trial ship maneuvering steps are as follows: 1. The device is powered on and the display screen of the trial ship maneuvering plan lights up.

[0081] 2. The operator rotates the knob of the trial steering module to adjust the pre-control command of the rudder angle, pushes the crossbar of the trial steering clock module forward and backward to adjust the pre-control command of the ship speed, and adjusts the time delay by pressing the button.

[0082] 3. After the parameters are set, press the "On / Off" button to turn on the trial maneuvering function. After being processed by the trial maneuvering function module, the collision avoidance effect under the current pre-maneuvering command conditions is dynamically presented on the trial maneuvering plan display.

[0083] 4. If the collision avoidance plan is feasible, the trial maneuvering ends; if the collision avoidance plan is not feasible, press the "Reset" button and repeat steps 2 and 3 until a feasible collision avoidance plan is obtained.

[0084] like Figure 6-Figure 10 After the operator inputs the corresponding parameters through the pre-manipulation control panel, the radar trial ship maneuvering function dynamically simulates the motion trajectory of the ship at times T0-T4. According to the effect shown in the figure, the collision avoidance plan at this time can enable the ship to perform avoidance operations safely.

[0085] like Fig.11 As shown, the predicted position of the target ship 2 appears within the closest encounter distance circle of the ship at the same time. At this time, the collision avoidance plan of the ship cannot enable the ship to safely avoid the target ship 2. It is necessary to adjust the parameter value through the pre-manipulation control panel until the collision avoidance plan is safe and feasible.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A radar test ship maneuvering method based on adaptive identification of ship dynamic characteristics, characterized in that: The following steps are involved: The ship's historical navigation data is modeled by machine learning identification, so as to construct a ship maneuvering motion model, wherein the ship maneuvering motion model is used to predict the ship's motion parameters and trajectory; the ship's historical navigation data includes the ship's position, heading, speed, rudder angle, propeller speed and torque, and the ship's motion parameters include the ship's turning radius, acceleration, bow turning rate and lag distance; the ship's maneuvering motion model includes a one-dimensional convolution layer, an LSTM layer, a multi-head attention mechanism, a first fully connected layer and a second fully connected layer, wherein the ship's historical navigation data is processed by the one-dimensional convolution layer to generate input data, and the input data is input into the LSTM layer, and the LSTM layer introduces a Dropout mechanism, and then the output of the LSTM layer is dynamically weighted according to the importance of the ship's maneuvering motion data through the multi-head attention mechanism, and the data processed by the multi-head attention mechanism is input into the first fully connected layer, and the first fully connected layer introduces a Dropout mechanism, and finally the information processed by the first fully connected layer is mapped to the prediction output space through the second fully connected layer; Based on the trained and optimized ship maneuvering motion model, the predicted ship motion parameters and trajectory are obtained, and the predicted ship motion parameters and trajectory are used to simulate the dynamic characteristics of the ship during the radar trial maneuvering process, providing a basis for the verification of the radar trial maneuvering collision avoidance plan; The ship trial maneuvering instruction is obtained, and the feasibility of the radar trial maneuvering plan is determined according to the radar trial maneuvering ship collision avoidance safety threshold in combination with the predicted ship motion parameters and trajectory. The radar trial maneuvering ship collision avoidance safety threshold includes the closest encounter distance between the own ship and the target ship.

2. According to claim 1, a radar trial ship maneuvering method based on adaptive identification of ship dynamic characteristics is characterized in that: It also includes a dynamic display of the ship's predicted trajectory and a visual display of the results of the ship's current collision avoidance plan.

3. The radar test ship maneuvering method based on adaptive identification of ship dynamic characteristics according to claim 1, characterized in that: The training optimization steps of the ship maneuvering motion model include: Preprocess the historical ship navigation data to construct training data sets, verification data sets, and test data sets; Training the ship maneuvering motion model based on a training data set; Performing hyperparameter tuning on the trained ship maneuvering motion model based on the validation data set; The prediction performance of the optimized ship maneuvering motion model is verified based on the test data set.

4. The radar test ship maneuvering method based on adaptive identification of ship dynamic characteristics according to claim 3 is characterized in that: The training and optimization step of the ship maneuvering motion model also includes: Automatically update and optimize the model based on real-time updated ship data.

5. A radar test ship maneuvering device based on adaptive identification of ship dynamic characteristics, used to implement the method according to any one of claims 1 to 4, characterized in that: include: A ship maneuvering motion modeling module, which uses machine learning to identify historical ship navigation data to model the ship, thereby constructing a ship maneuvering motion model, and is used to output predicted ship motion parameters and trajectories based on the constructed ship maneuvering motion model; the historical ship navigation data includes the ship's position, speed and heading, and the ship's motion parameters include the ship's turning radius, acceleration and bow turning rate; A pre-manipulation control panel module, which is used to obtain pre-manipulation commands on the one hand, and to display ship motion parameters predicted by identification modeling on the other hand; The dynamic generation module of the trial maneuvering plan is used to obtain the ship's trial maneuvering instructions, combine the predicted ship motion parameters and trajectory, and determine the feasibility of the trial maneuvering plan according to the collision avoidance safety threshold. On the other hand, it is used to dynamically display the predicted trajectory of the ship and visualize the execution results of the ship's current collision avoidance plan.

6. The radar test ship maneuvering device based on adaptive identification of ship dynamic characteristics according to claim 5, characterized in that: The dynamic generation module of the trial ship maneuvering plan includes a trial ship maneuvering function module and a trial ship maneuvering plan display; The trial ship maneuvering function module predicts the future motion data of the ship and calculates the corresponding motion trajectory based on the obtained ship maneuvering motion model and the ship pre-maneuvering command input by the pre-maneuvering control panel; The trial ship maneuvering plan display dynamically displays the ship motion trajectory under different pre-manipulation control command conditions, so that the radar trial ship maneuvering plan is accurately presented on the display screen.

7. The radar test ship maneuvering device based on adaptive identification of ship dynamic characteristics according to claim 5, characterized in that: When implementing radar trial ship maneuvering, the operator generates an input signal through the operation area on the pre-manipulation control panel. The input signal is captured by the circuit of the pre-manipulation control panel, converted into a digital signal, and then transmitted to the trial ship maneuvering function module through the bus after being processed by signal amplification, filtering, and analog-to-digital conversion steps. After receiving the signal from the pre-manipulation control panel, the trial ship maneuvering function module determines the operation to be performed according to the preset program and logical judgment, and dynamically displays it through the trial ship maneuvering plan display.

Images