Unmanned ship prediction tracking control method and system based on GRU disturbance observer

Abstract

The invention relates to an unmanned ship prediction tracking control method and system based on a GRU disturbance observer, and belongs to the field of ocean intelligent equipment control. The method comprises the following steps: constructing an unmanned ship kinematics model and an unmanned ship dynamics model; transmitting sensor data to the unmanned ship kinematics model and the unmanned ship dynamics model, and calculating an actual disturbance value; and obtaining an unmanned ship linearization model according to the constructed GRU disturbance observer model, and obtaining a target control quantity for the unmanned ship linearization model through a model prediction control strategy. The method solves the problems that the accuracy of tracking the expected path by the unmanned ship is reduced and the course error is increased due to low prediction precision of model prediction control.

Description

Technical Field

[0001] This application relates to the field of marine intelligent equipment control technology, and in particular to a predictive tracking control method and system for unmanned vessels based on a GRU disturbance observer. Background Technology

[0002] With the development of the marine economy and the construction of smart ocean, unmanned surface vessels (USVs) are increasingly widely used in fields such as marine surveying, environmental monitoring, and maritime patrol. One of their core functions is to achieve accurate and stable tracking of desired paths or trajectories by USVs.

[0003] In existing technologies, common tracking control methods include PID control, Active Disturbance Rejection Control (ADRC), and Model Predictive Control (MPC). Among these, MPC is favored because it can explicitly handle system constraints and multi-objective optimization. However, the performance of MPC is highly dependent on the accuracy of the predictive model used. The dynamic models of unmanned surface vessels (USVs) are usually quite complex, and in actual operation, they are strongly affected by external environmental factors such as wind, waves, and currents, as well as the hydrodynamic characteristics of the vessel itself. These effects can be regarded as "disturbances" acting on the system. Because these disturbances are time-varying and uncertain, it is difficult to establish accurate mathematical models, leading to large errors in MPC based on approximate models.

[0004] To address the aforementioned issues, existing technologies propose using radial basis function neural networks (RBF NNs) as disturbance observers to estimate disturbance values ​​in real time and use these estimates to compensate the system, thereby simplifying the dynamic model and improving the prediction accuracy of model predictive control. This method effectively improves control performance and reduces manual parameter tuning. However, it still has the following shortcomings: (1) Radial basis function neural networks cannot capture the dynamic characteristics of disturbances evolving over time, such as the periodicity of waves and the suddenness of gusts, which limits their estimation accuracy; (2) The motion of unmanned vessels and the disturbances they experience have a strong temporal correlation. Relying solely on the current state information for disturbance estimation ignores the value of historical information, which may lead to a delayed response to sudden disturbances; (3) Existing disturbance observers typically rely only on the motion and thrust data of the vessel itself, failing to fully utilize the direct environmental information provided by external environmental sensors, such as anemometers and wave meters. Summary of the Invention

[0005] This application provides a predictive tracking control method and system for unmanned vessels based on a GRU disturbance observer, which can solve the problem that the accuracy of unmanned vessels in tracking the desired path and the heading error increase due to the low prediction accuracy of model predictive control.

[0006] To achieve the above objectives, according to the first aspect of this application, a predictive tracking control method for unmanned surface vessels based on a GRU disturbance observer is provided, specifically including the following steps: S1. Construct the kinematic model and dynamic model of the unmanned vessel; S2, in each control cycle The system acquires IMU sensor data at the current moment and transmits it to the unmanned surface vessel's kinematics model and dynamics model, respectively, to obtain the current forward velocity. Differential forward velocity and the actual disturbance value of the unmanned vessel propulsion module and ; S3. Construct a GRU perturbation observer, and input vector The input vector is fed into the GRU perturbation observer for training. Including forward speed and differential forward velocity After training, a pre-trained GRU perturbation observer model is obtained, which outputs perturbation estimates. and ; S4. Based on the disturbance estimate and Calculate the compensation thrust ; estimate the disturbance and Replace actual disturbance value and The linearized model of the unmanned vessel is obtained; the target control quantity is obtained through the model predictive control strategy of the unmanned vessel linearized model. ; to control the target quantity With compensating thrust The summation yields the final target control output sent to the unmanned vessel propulsion module. ; S5, in the next control cycle Repeat steps S2-S4 to achieve closed-loop control.

[0007] Furthermore, the construction of the unmanned vessel kinematic model in S1 includes the following steps: The unmanned ship's location is Derivative velocity Rotation speed forward speed lateral velocity and heading angle The kinematic model of the unmanned vessel is constructed using the following formula:

[0008]

[0009]

[0010] in, This is the yaw rate.

[0011] Furthermore, the construction of the unmanned vessel dynamics model in S1 includes the following steps: Set the mass of the unmanned vessel Moment of inertia Spacing between left and right thrusters Left thruster thrust And right thruster reasoning Set unknown disturbances, including and ,in Resistance in the direction of forward movement, Resistance in the forward direction The torque generated; Assuming the unmanned vessel's driving state is underactuated, construct the unmanned vessel's dynamic model:

[0012]

[0013] in, Let the differential forward velocity be... This represents the rotational speed.

[0014] Furthermore, step S2 includes the following steps: During the control cycle Within the current timeframe, acquire IMU sensor data, including derivative velocity. Heading angle and rotational speed ; Derivative speed and heading angle Substituting the actual value into the kinematic model of the unmanned vessel, we obtain the forward velocity. and differential forward velocity ; Get rotation speed and unmanned ship quality Based on the unmanned vessel dynamics model, the actual disturbance value of the unmanned vessel propulsion module is calculated using the following formula. and :

[0015]

[0016] in, To control the cycle The actual thrust generated by the inner left thruster To control the cycle The actual thrust generated by the inner right thruster.

[0017] Furthermore, the input vector Including hull condition and environmental information, the calculation is performed using the following formula:

[0018] in, To control the cycle The ambient wind speed measured in real time inside, To control the cycle The wave height measured in real time, and the ambient wind speed. and Wave High All data is acquired by environmental sensors; if no environmental sensors are available, then... , .

[0019] Furthermore, the GRU perturbation observer construction process of S3 includes the following steps: Let the dimension of the input layer be equal to the input vector. dimensionality; Set up a hidden layer, which includes one or more GRU units, each GRU unit containing an update gate. Reset door Candidate hidden state and current hidden state ; The output layer is set as a fully connected layer, and the input of the output layer is the last-stage GRU unit in the current control cycle. Hidden state of output Output layer mapping generates perturbation estimates. The disturbance estimate is calculated using the following formula:

[0020] in, This is the output layer weight matrix. It is the bias vector; Update Gate The hidden state at the previous time step is determined by the following formula. Information that needs to be retained:

[0021] in, To update the door The input layer weight matrix is ​​used to weight the current input vector. Mapped to a linear combination of update gates, To reset the door The hidden layer weight matrix is ​​used to store the hidden state from the previous time step. A linear combination mapped to reset doors. To reset the door The bias vector is used to adjust the reset gate. The threshold of the activation function; Reset door The hidden state at the previous time step is determined by the following formula. Ignored information:

[0022] in, To reset the door The input layer weight matrix; Candidate hidden state Based on input vector and reset door The information is used to calculate the new candidate state using the following formula:

[0023] in, For element-wise multiplication; Current hidden state Combined with the updated gate and candidate hidden state Information, through formula Generate the final hidden state.

[0024] Furthermore, the training process of the GRU perturbation observer in S3 includes the following steps: In the simulation environment or during the actual ship testing phase, the input vector and their corresponding actual disturbance values The data is input into the GRU network structure; Set a loss function, which is as follows:

[0025] The network weights are updated by minimizing the loss function using the backpropagation algorithm. Training is complete, resulting in a pre-trained GRU perturbation observer model, which outputs perturbation estimates. and .

[0026] Furthermore, step S4 includes the following steps: The disturbance estimate output by the constructed GRU disturbance observer model and The compensating thrust is calculated using the following formula:

[0027]

[0028]

[0029] in, To control the cycle Internal disturbance compensation thrust vector To control the cycle The equivalent compensating thrust generated by the internal disturbance on the left thruster. To control the cycle The equivalent compensating thrust generated by the internal cancellation disturbance on the right thruster; The disturbance estimate and Actual perturbation value of the alternative dynamics model and The linearized model of the unmanned vessel is obtained through the following formula:

[0030]

[0031] In the current control cycle Inside, and It is a known constant; Based on the linearized model of the unmanned vessel, the target control quantity is obtained through model predictive control strategy. ; The overall target control output sent from the unmanned surface vessel to the thrusters is obtained by the following formula:

[0032]

[0033]

[0034] in, In the control cycle The overall target control output vector of the unmanned surface vessel propulsion module, In the control cycle The final target thrust command is sent internally to the left thruster. In the control cycle The final target thrust command is sent internally to the right thruster. In the control cycle The target control quantity sent to the left thruster. In the control cycle The target control quantity is sent to the right thruster.

[0035] Furthermore, the model predictive control strategy includes the following steps: The linear model of the unmanned vessel is discretized to obtain discrete state equations; The desired forward speed is obtained through path planning algorithms. Desired heading angle Predicting step size and control step size Build a predictive model; Set the quadratic performance function Minimize the prediction step size using the following formula Tracking error and control quantity changes within:

[0036] in, To determine the error between the predicted output and the expected output, Here is the error weight matrix. To control the increase in quantity, To control the incremental weight matrix; Satisfying the thruster inference constraints Under the conditions, obtain the future The optimal control quantity sequence at each time point; Select the first control variable in the optimal control variable sequence as the target control variable. .

[0037] To achieve the above objectives, according to a second aspect of this application, an unmanned vessel predictive tracking control system based on a GRU disturbance observer is also provided. The system includes a data acquisition module, a GRU disturbance observer module, a compensation calculation module, a model predictive control module, an output synthesis module, and a control command execution module. The data acquisition module obtains the input vector. And transmit it to the GRU perturbation observer module; The GRU perturbation observer module stores pre-trained GRU perturbation observers, which are used to process input vectors. Process the data and output the disturbance estimate. and To the compensation calculation module; The compensation calculation module calculates based on the disturbance estimate. and Calculate the compensation thrust The disturbance estimate and Replace actual disturbance value and The linearized model of the unmanned vessel is obtained; the target control quantity is obtained through the model predictive control strategy of the unmanned vessel linearized model. ; to compensate for thrust and target control quantity Transmitted to the model prediction control module; The model predictive control module will compensate for thrust. and target control quantity And transmit it to the output synthesis module; The output synthesis module will compensate for thrust. and target control quantity The sum is the total target control output. The overall target control output Transmitted to the control command execution module; The control command execution module transmits the final overall target control output to the unmanned vessel propulsion unit for execution.

[0038] The technical effects of this application are as follows: (1) The powerful temporal modeling capability of the GRU network enables it to effectively capture the dynamic changes of disturbances. Compared with the existing static RBF NN, it has higher estimation accuracy for periodic and sudden disturbances, especially under complex sea conditions; (2) This application brings more effective compensation through more accurate disturbance estimation, which greatly reduces the sensitivity of the system to model mismatch and external interference, and makes the tracking process more stable and reliable; (3) This application performs model predictive control prediction based on the linear model after accurate compensation, which ensures the accuracy of the prediction, so that the unmanned vessel can track the desired path or trajectory more accurately and reduce position and heading errors; (4) By fusing environmental sensor information, this application enables the system to predict environmental changes and adjust the control strategy in advance, further improving the adaptability in dynamic environments; (5) The entire disturbance observation and compensation process of this application is automated, avoiding the complex parameter tuning process in traditional methods and reducing the difficulty of system deployment and maintenance. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is a flowchart of an unmanned surface vessel predictive tracking control method based on a GRU disturbance observer, according to an embodiment of this application. Figure 2 This is a structural diagram of an unmanned surface vessel predictive tracking control system based on a GRU disturbance observer, according to an embodiment of this application. Detailed Implementation

[0041] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0042] This application provides a predictive tracking control method for unmanned surface vessels based on a GRU disturbance observer, the method comprising the following steps: S1. Construct the kinematic model and dynamic model of the unmanned vessel; S2. Transmit the IMU sensor data to the unmanned vessel's kinematics model and dynamics model, and calculate the actual disturbance value; S3. Based on the constructed GRU disturbance observer model, obtain the linearized model of the unmanned vessel, and obtain the target control quantity from the linearized model of the unmanned vessel through the model predictive control strategy.

[0043] Furthermore, S1 includes the following steps: Establishing unmanned ships in a geodetic coordinate system With the ship's coordinate system The kinematic relationship between them; let the position of the unmanned vessel be Derivative velocity Rotation speed forward speed lateral velocity and heading angle The kinematic model of the unmanned vessel is constructed using the following formula:

[0044]

[0045]

[0046] in, Yaw angular velocity; Set the mass of the unmanned vessel Moment of inertia Spacing between left and right thrusters Left thruster thrust And right thruster reasoning Set unknown disturbances, including and ,in Resistance in the direction of forward movement, Resistance in the forward direction The torque generated; The unmanned vessel is assumed to be in an underactuated state, and lateral resistance is ignored. Construct an unmanned vessel dynamics model:

[0047]

[0048] in, Let the differential forward velocity be... This represents the rotational speed.

[0049] Furthermore, S2 includes the following steps: During the control cycle Within the current timeframe, acquire IMU sensor data, including derivative velocity. Heading angle and rotational speed ; Derivative speed and heading angle Substituting the actual value into the kinematic model of the unmanned vessel, we obtain the forward velocity. and differential forward velocity ; Get rotation speed and unmanned ship quality Based on the unmanned vessel dynamics model, the actual disturbance value of the unmanned vessel propulsion module is calculated using the following formula. and :

[0050]

[0051] in, To control the cycle The actual thrust generated by the inner left thruster To control the cycle The actual thrust generated by the inner right thruster.

[0052] Furthermore, S3 includes the following steps: The input vector for constructing the GRU disturbance observer model is based on hull condition and environmental information. ; Construct a GRU network structure, including an input layer, a hidden layer, and an output layer, wherein the hidden layer includes one or more GRU units; In the simulation environment or during the actual ship testing phase, the input vector The actual perturbation values ​​are input into the GRU network structure and trained to obtain a pre-trained GRU perturbation observer model.

[0053] Constructing input vectors As shown in the following formula, the input vector not only includes the hull state but also incorporates environmental information:

[0054] in, To control the cycle The ambient wind speed measured in real time inside, To control the cycle The wave height measured in real time, the and All data is acquired by environmental sensors; if no environmental sensors are available, then... , .

[0055] Building a GRU network structure includes the following steps: Let the dimension of the input layer be equal to the input vector. dimensionality; Set up a hidden layer, which includes one or more GRU units, each GRU unit containing an update gate. Reset door Candidate hidden state and current hidden state ; The output layer is set as a fully connected layer, and the input of the output layer is the last-stage GRU unit in the current control cycle. Hidden state of output Output layer mapping generates perturbation estimates. The disturbance estimate is calculated using the following formula:

[0056] in, This is the output layer weight matrix. It is the bias vector; Update Gate The hidden state at the previous time step is determined by the following formula. Information that needs to be retained:

[0057] in, To update the door The input layer weight matrix is ​​used to weight the current input vector. Mapped to a linear combination of update gates, To reset the door The hidden layer weight matrix is ​​used to store the hidden state from the previous time step. A linear combination mapped to reset doors. To reset the door The bias vector is used to adjust the reset gate. The threshold of the activation function; Reset door The hidden state at the previous time step is determined by the following formula. Ignored information:

[0058] in, To reset the door The input layer weight matrix is ​​used to weight the current input vector. A linear combination of maps to reset gates is used to determine the hidden state of the previous time step. The degree to which candidate hidden states are "forgotten" or "ignored" when calculating the hidden states; Candidate hidden state Based on input vector and reset door The information is used to calculate the new candidate state using the following formula:

[0059] in, For element-wise multiplication; Current hidden state Combined with the updated gate and candidate hidden state Information, through formula Generate the final hidden state.

[0060] Furthermore, training the GRU network structure includes the following steps: In the simulation environment or during the actual ship testing phase, the input vector and their corresponding actual disturbance values The data is input into the GRU network structure; Set a loss function, which is as follows:

[0061] The network weights are updated by minimizing the loss function using the backpropagation algorithm. Training is complete, resulting in a pre-trained GRU perturbation observer model, which outputs perturbation estimates. and .

[0062] Furthermore, S4 includes the following steps: The disturbance estimate output by the constructed GRU disturbance observer model and The compensating thrust is calculated using the following formula:

[0063]

[0064]

[0065] in, To control the cycle The internal disturbance compensation thrust vector is generated from the disturbance estimate. To control the cycle The equivalent compensating thrust generated by the internal disturbance on the left thruster. To control the cycle The equivalent compensating thrust generated by the internal cancellation disturbance on the right thruster; The disturbance estimate and Actual perturbation value of the alternative dynamics model and The linearized model of the unmanned vessel is obtained through the following formula:

[0066]

[0067] In the current control cycle Inside, and These are known constants, thus making the model an exact linearized model that can be used for model predictive control.

[0068] Based on the linearized model of the unmanned vessel, the target control quantity is obtained through model predictive control strategy. ; The overall target control output sent from the unmanned surface vessel to the thrusters is obtained by the following formula:

[0069]

[0070]

[0071] in, In the control cycle The overall target control output vector of the unmanned surface vessel propulsion module, In the control cycle The final target thrust command is sent internally to the left thruster. In the control cycle The final target thrust command is sent internally to the right thruster. In the control cycle The target control quantity sent to the left thruster. In the control cycle The target control quantity is sent to the right thruster.

[0072] Furthermore, the model predictive control strategy includes the following steps: The linear model of the unmanned vessel is discretized to obtain discrete state equations; Based on the current discrete state, the desired forward speed is obtained through a path planning algorithm. Desired heading angle Predicting step size and control step size Build a predictive model; Define quadratic performance function Minimize the prediction step size using the following formula Tracking error and control quantity changes within:

[0073] in, To determine the error between the predicted output and the expected output, Here is the error weight matrix. To control the increase in quantity, To control the incremental weight matrix; Solve the following optimization problem: Given the thruster inference constraints... Under the conditions, obtain the future The optimal control quantity sequence at each time point; Select the first control variable in the optimal control variable sequence as the target control variable. .

[0074] This application also provides an unmanned vessel predictive tracking control system based on a GRU disturbance observer, including a data acquisition module, a GRU disturbance observer module, a compensation calculation module, a model predictive control module, an output synthesis module, and a control command execution module; The data acquisition module obtains the input vector. And transmit it to the GRU perturbation observer module; The GRU perturbation observer module stores pre-trained GRU perturbation observers, which are used to process input vectors. Process the data and output the disturbance estimate. and To the compensation calculation module; The compensation calculation module calculates based on the disturbance estimate. and Calculate the compensation thrust The disturbance estimate and Replace actual disturbance value and The linearized model of the unmanned vessel is obtained; the target control quantity is obtained through the model predictive control strategy of the unmanned vessel linearized model. ; to compensate for thrust and target control quantity Transmitted to the model prediction control module; The model predictive control module will compensate for thrust. and target control quantity And transmit it to the output synthesis module; The output synthesis module will compensate for thrust. and target control quantity The sum is the total target control output. The overall target control output Transmitted to the control command execution module; The control command execution module transmits the final overall target control output to the unmanned vessel propulsion unit for execution.

[0075] In one specific embodiment, a ship with a mass of Moment of inertia is The distance between the thrusters is The dual-propeller unmanned surface vessel (USV) performs a straight-line tracking mission in sea areas with regular waves and random gusts of wind. The USV is equipped with a GPS / IMU integrated navigation module and left / right thrusters (thrust range...). Anemometer and wave meter; control cycle is The GRU network includes one hidden layer with 16 neurons; the prediction step size for model prediction control. Control step size .

[0076] The control flow in this embodiment includes the following steps: exist At a certain time, the system initializes, and the GRU network loads the pre-trained weights; The data acquisition module obtains the current status, including location. ,speed ,course Rotation speed forward speed Left thrust Right thrust Wind speed and Wave High ; The actual disturbance value was calculated based on the kinematic and dynamic models of the unmanned surface vessel (USV). , ; Construct the input vector The input is fed into the GRU perturbation observer module; The GRU network calculates and outputs an estimated disturbance value. , ; The compensation calculation module calculates the disturbance compensation output according to the formula:

[0077]

[0078] Model predictive control module utilizes and Simplify the dynamic model and combine the desired velocity and expected course Solve the model predictive control problem to obtain the target control quantity at the current time. ; The output synthesis module calculates the final overall target control output:

[0079] Among them, limited by Therefore Pick ;

[0080] Among them, limited by Therefore Pick ; The control instruction execution module will and The instructions are sent to the thruster module; System entry At any given time, repeat the above process; the hidden state of the GRU network. It will be passed to the next moment for calculation. This enables the memorization of temporal information.

[0081] The technical effects of this application are as follows: (1) The powerful temporal modeling capability of the GRU network enables it to effectively capture the dynamic changes of disturbances. Compared with the existing static RBF NN, it has higher estimation accuracy for periodic and sudden disturbances, especially under complex sea conditions; (2) This application brings more effective compensation through more accurate disturbance estimation, which greatly reduces the sensitivity of the system to model mismatch and external interference, and makes the tracking process more stable and reliable; (3) This application performs model predictive control prediction based on the linear model after accurate compensation, which ensures the accuracy of the prediction, so that the unmanned vessel can track the desired path or trajectory more accurately and reduce position and heading errors; (4) By fusing environmental sensor information, this application enables the system to predict environmental changes and adjust the control strategy in advance, further improving the adaptability in dynamic environments; (5) The entire disturbance observation and compensation process of this application is automated, avoiding the complex parameter tuning process in traditional methods and reducing the difficulty of system deployment and maintenance.

[0082] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps (steps of the method). The storage medium is, for example, ROM / RAM, magnetic disk, optical disk, etc.

[0083] This application also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the various steps of the aforementioned embodiments of the unmanned vessel predictive tracking control method and system based on GRU disturbance observers, and can achieve the same beneficial effects as the aforementioned embodiments of the unmanned vessel predictive tracking control method and system based on GRU disturbance observers. To avoid repetition, they will not be described again here.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0085] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can be physically included separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0086] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute some steps of the transmission and reception methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0087] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A predictive tracking control method for unmanned surface vessels based on a GRU disturbance observer, characterized in that, Specifically, it includes the following steps: S1. Construct the kinematic model and dynamic model of the unmanned vessel; S2, in each control cycle The system acquires IMU sensor data at the current moment and transmits it to the unmanned surface vessel's kinematics model and dynamics model, respectively, to obtain the current forward velocity. Differential forward velocity and the actual disturbance value of the unmanned vessel propulsion module and ; S3. Construct a GRU perturbation observer, and input vector The input vector is fed into the GRU perturbation observer for training. Including forward speed and differential forward velocity ; After training, a pre-trained GRU perturbation observer model is obtained, which outputs perturbation estimates. and ; S4. Based on the disturbance estimate and Calculate the compensation thrust ; estimate the disturbance and Replace actual disturbance value and The linearized model of the unmanned vessel is obtained; the target control quantity is obtained through the model predictive control strategy of the unmanned vessel linearized model. ; to control the target quantity With compensating thrust The summation yields the final target control output sent to the unmanned vessel propulsion module. ; S5, in the next control cycle Repeat steps S2-S4 to achieve closed-loop control.

2. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 1, characterized in that, The construction of the unmanned vessel kinematic model in S1 includes the following steps: The unmanned ship's location is Derivative velocity Rotation speed forward speed lateral velocity and heading angle The kinematic model of the unmanned vessel is constructed using the following formula: in, This is the yaw rate.

3. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 2, characterized in that, The construction of the unmanned vessel dynamics model in S1 includes the following steps: Set the mass of the unmanned vessel Moment of inertia Spacing between left and right thrusters Left thruster thrust And right thruster reasoning ; Set unknown disturbances, including and ,in Resistance in the direction of forward movement, Resistance in the forward direction The torque generated; Assuming the unmanned vessel's driving state is underactuated, construct the unmanned vessel's dynamic model: in, Let the differential forward velocity be... This represents the rotational speed.

4. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 3, characterized in that, S2 includes the following steps: During the control cycle Within the current timeframe, acquire IMU sensor data, including derivative velocity. Heading angle and rotational speed ; Derivative speed and heading angle Substituting the actual value into the kinematic model of the unmanned vessel, we obtain the forward velocity. and differential forward velocity ; Get rotation speed and unmanned ship quality Based on the unmanned vessel dynamics model, the actual disturbance value of the unmanned vessel propulsion module is calculated using the following formula. and : in, To control the cycle The actual thrust generated by the inner left thruster To control the cycle The actual thrust generated by the inner right thruster.

5. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 4, characterized in that, The input vector Including hull condition and environmental information, the calculation is performed using the following formula: in, To control the cycle The ambient wind speed measured in real time inside, To control the cycle The wave height measured in real time, and the ambient wind speed. and Wave High All data is acquired by environmental sensors; if no environmental sensors are available, then... , .

6. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 5, characterized in that, The construction process of the GRU perturbation observer in S3 includes the following steps: Let the dimension of the input layer be equal to the input vector. dimensionality; Set up a hidden layer, which includes one or more GRU units, each GRU unit containing an update gate. Reset door Candidate hidden state and current hidden state ; The output layer is set as a fully connected layer, and the input of the output layer is the last-stage GRU unit in the current control cycle. Hidden state of output ; Output layer mapping generates perturbation estimates The disturbance estimate is calculated using the following formula: in, This is the output layer weight matrix. It is the bias vector; Update Gate The hidden state at the previous time step is determined by the following formula. Information that needs to be retained: in, To update the door The input layer weight matrix is ​​used to weight the current input vector. Mapped to a linear combination of update gates, To reset the door The hidden layer weight matrix is ​​used to store the hidden state from the previous time step. A linear combination mapped to reset doors. To reset the door The bias vector is used to adjust the reset gate. The threshold of the activation function; Reset door The hidden state at the previous time step is determined by the following formula. Ignored information: in, To reset the door The input layer weight matrix; Candidate hidden state Based on input vector and reset door The information is used to calculate the new candidate state using the following formula: in, Candidate hidden state The input layer weight matrix, Candidate hidden state Hidden layer weight matrix, Candidate hidden state The bias vector, For element-wise multiplication; Current hidden state Combined with the updated gate and candidate hidden state Information, through formula Generate the final hidden state.

7. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 6, characterized in that, The training process of the GRU perturbation observer in S3 includes the following steps: In the simulation environment or during the actual ship testing phase, the input vector and their corresponding actual disturbance values The data is input into the GRU network structure; Set a loss function, which is as follows: The network weights are updated by minimizing the loss function using the backpropagation algorithm. Training is complete, resulting in a pre-trained GRU perturbation observer model, which outputs perturbation estimates. and .

8. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to claim 7, characterized in that, S4 includes the following steps: The disturbance estimate output by the constructed GRU disturbance observer model and The compensating thrust is calculated using the following formula: in, To control the cycle Internal disturbance compensation thrust vector To control the cycle The equivalent compensating thrust generated by the internal disturbance on the left thruster. To control the cycle The equivalent compensating thrust generated by the internal cancellation disturbance on the right thruster; The disturbance estimate and Actual perturbation value of the alternative dynamics model and The linearized model of the unmanned vessel is obtained through the following formula: In the current control cycle Inside, and It is a known constant; Based on the linearized model of the unmanned vessel, the target control quantity is obtained through model predictive control strategy. ; The overall target control output sent from the unmanned surface vessel to the thrusters is obtained by the following formula: in, In the control cycle The overall target control output vector of the unmanned surface vessel propulsion module, In the control cycle The final target thrust command is sent internally to the left thruster. In the control cycle The final target thrust command is sent internally to the right thruster. In the control cycle The target control quantity sent to the left thruster. In the control cycle The target control quantity is sent to the right thruster.

9. The unmanned surface vessel predictive tracking control method based on a GRU disturbance observer according to any one of claims 1-8, characterized in that, The model predictive control strategy includes the following steps: The linear model of the unmanned vessel is discretized to obtain discrete state equations; The desired forward speed is obtained through path planning algorithms. Desired heading angle Predicting step size and control step size Build a predictive model; Set the quadratic performance function Minimize the prediction step size using the following formula Tracking error and control quantity changes within: in, To determine the error between the predicted output and the expected output, Here is the error weight matrix. To control the increase in quantity, To control the incremental weight matrix; Satisfying the thruster inference constraints Under the conditions, obtain the future The optimal control quantity sequence at each time point; Select the first control variable in the optimal control variable sequence as the target control variable. .

10. A predictive tracking control system for unmanned surface vessels based on a GRU disturbance observer, characterized in that, The system includes a data acquisition module, a GRU disturbance observer module, a compensation calculation module, a model prediction control module, an output synthesis module, and a control command execution module; The data acquisition module obtains the input vector. And transmit it to the GRU perturbation observer module; The GRU perturbation observer module stores pre-trained GRU perturbation observers, which are used to process input vectors. Process the data and output the disturbance estimate. and To the compensation calculation module; The compensation calculation module calculates based on the disturbance estimate. and Calculate the compensation thrust The disturbance estimate and Replace actual disturbance value and The linearized model of the unmanned vessel is obtained; the target control quantity is obtained through the model predictive control strategy of the unmanned vessel linearized model. ; to compensate for thrust and target control quantity Transmitted to the model prediction control module; The model predictive control module will compensate for thrust. and target control quantity And transmit it to the output synthesis module; The output synthesis module will compensate for thrust. and target control quantity The sum is the total target control output. The overall target control output Transmitted to the control command execution module; The control command execution module transmits the final overall target control output to the unmanned vessel propulsion unit for execution.

Images