CN-121979191-A - Unmanned ship prediction tracking control method and system based on GRU disturbance observer

CN121979191ACN 121979191 ACN121979191 ACN 121979191ACN-121979191-A

Abstract

The application relates to an unmanned ship predictive 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 steps of constructing an unmanned ship kinematic model and an unmanned ship dynamics model, transmitting sensor data into the unmanned ship kinematic model and the unmanned ship dynamics model, calculating an actual disturbance value, obtaining an unmanned ship linearization model according to the constructed GRU disturbance observer model, and obtaining target control quantity for the unmanned ship linearization model through a model predictive control strategy. The method solves the problems of reduced accuracy of the unmanned ship tracking expected path and increased course error caused by low prediction accuracy of model prediction control.

Inventors

WANG JIEKAI
ZHOU GUANGHAI
JI JUNPING
HUANG ZHIZHOU
CHEN HANLONG

Assignees

广州南方测绘科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251212

Claims (10)

1. The unmanned ship prediction tracking control method based on the GRU disturbance observer is characterized by comprising the following steps of: S1, constructing an unmanned ship kinematic model and an unmanned ship kinematic model; s2, in each control period Acquiring IMU sensor data at the current moment, and respectively transmitting the IMU sensor data to an unmanned ship kinematics model and an unmanned ship dynamics model to respectively obtain the advancing speed at the current moment Differential forward speed And actual disturbance value of unmanned ship propeller module And ; S3, constructing a GRU disturbance observer, and inputting vectors Input to the GRU disturbance observer for training, the input vector Including forward speed And differential forward speed Obtaining a pre-trained GRU disturbance observer model after training is completed, and outputting disturbance estimated values And ; S4, according to the disturbance estimated value And Calculating the compensation thrust Disturbance estimation value And Instead of actual disturbance values And Obtaining a linearization model of the unmanned ship, wherein the linearization model of the unmanned ship obtains a target control amount through a model predictive control strategy Controlling the target amount And compensating for thrust Adding to obtain a total target control output which is finally sent to the unmanned ship propeller module ; S5, in the next control period And S2-S4 is repeatedly executed, so that closed-loop control is realized.
2. The unmanned ship predictive tracking control method based on the GRU disturbance observer according to claim 1, wherein the step of constructing the unmanned ship kinematic model of S1 comprises the following steps: the unmanned ship is positioned as Speed of derivative Rotational speed Speed of forward travel Transverse velocity And heading angle The unmanned ship kinematics model is constructed by the following formula: Wherein, the Is yaw rate.
3. The unmanned ship predictive tracking control method based on the GRU disturbance observer according to claim 2, wherein the constructing the unmanned ship dynamics model of S1 comprises the following steps: Setting the mass of an unmanned ship Moment of inertia Distance between left and right propellers Thrust of left propeller And right propeller reasoning Setting unknown disturbance including And Wherein In order to provide a resistance in the forward direction, Resistance to the advancing direction The moment generated; The driving state of the unmanned ship is an underactuated state, and a unmanned ship dynamics model is constructed: Wherein, the In order to differentiate the forward speed of the vehicle, Is the rotational speed.
4. The unmanned ship predictive tracking control method based on the GRU disturbance observer according to claim 3, wherein the step S2 comprises the following steps: During the control period Acquiring IMU sensor data at the current moment, wherein the IMU sensor data comprises derivative speed Course angle And rotational speed ; The derivative speed And heading angle Substituting the actual value of (2) into the unmanned ship kinematic model to obtain the forward speed And differential forward speed ; Acquiring rotational speed And unmanned ship mass Combining the unmanned ship dynamics model, calculating the actual disturbance value of the unmanned ship propeller module through the following formula And : Wherein, the To control period The actual thrust generated by the inner left propeller, To control period The actual thrust generated by the inner right propeller.
5. The method for unmanned ship predictive tracking control based on GRU disturbance observer according to claim 4, wherein said input vector Including hull status and environmental information, calculated by: Wherein, the To control period The ambient wind speed measured in real time is measured, To control period Wave height measured in real time, the ambient wind speed Wave height Are acquired by environmental sensors, if no environmental sensor exists , 。
6. The unmanned ship predictive tracking control method based on the GRU disturbance observer according to claim 5, wherein the GRU disturbance observer constructing process of S3 comprises the following steps: let the dimension of the input layer equal to the input vector Dimension of (2); Providing a hidden layer comprising one or more GRU units, each GRU unit comprising an update door Reset gate Candidate hidden state And the current hidden state ; Setting an output layer as a full connection layer, wherein the input of the output layer is that the GRU unit of the last stage is in the current control period Hidden state of output Output layer mapping to generate disturbance estimation The disturbance estimate is calculated by: Wherein, the In order to output the layer weight matrix, Is a bias vector; Update door The hidden state of the previous moment is determined by Information that needs to be retained: Wherein, the To update the door For weighting matrix of input layer of (a) for inputting current vector Mapped to a linear combination of update gates, To reset the door For concealing the concealing state of the previous moment A linear combination mapped to the reset gate, To reset the door For adjusting reset gate Activating a threshold of the function; Reset gate The hidden state of the previous moment is determined by Ignored information: Wherein, the To reset the door Is a matrix of input layer weights; candidate hidden states Based on input vectors Reset gate Calculating a new candidate state by: Wherein, the For candidate hidden states Is used for the input layer weight matrix of the (c), For candidate hidden states The weight matrix of the hidden layer is used, For candidate hidden states Is set in the above-described state, For element-by-element multiplication; Current hidden state Combined updating door And candidate hidden states Through the formula A final hidden state is generated.
7. The unmanned ship predictive tracking control method based on the GRU disturbance observer according to claim 6, wherein the training process of the GRU disturbance observer of S3 comprises the following steps: in the simulation environment or real ship test stage, the input vector is input Actual disturbance value corresponding to the disturbance value Is input to the GRU network structure; Setting a loss function, wherein the loss function is as follows: Minimizing a loss function through a back propagation algorithm, and updating network weights; Training is completed, a pre-trained GRU disturbance observer model is obtained, and a disturbance estimated value is output And 。
8. The unmanned ship predictive tracking control method based on the GRU disturbance observer according to claim 7, wherein the step S4 comprises the following steps: Disturbance estimation value output according to constructed GRU disturbance observer model And The compensation thrust is calculated by: Wherein, the To control period The internal disturbance compensates for the thrust vector, To control period The inner offset disturbance produces an equivalent compensating thrust to the left propeller, To control period The inner offset disturbance generates equivalent compensation thrust to the right propeller; Disturbance estimation value And Actual disturbance value of surrogate dynamics model And The linearization model of the unmanned ship is obtained by the following formula: Wherein, in the current control period In the inner part of the inner part, And Is a known constant; Obtaining target control quantity through a model predictive control strategy according to the unmanned ship linearization model ; The total target control output sent to the propeller by the unmanned ship is obtained by: Wherein, the To be in control period The total target control output vector of the inner unmanned ship propeller module, To be in control period A final target thrust command internally sent to the left propeller, To be in control period A final target thrust command internally sent to the right propeller, To be in control period The target control amount sent to the left propeller, To be in control period The target control amount sent to the right propeller.
9. The unmanned ship predictive tracking control method based on a GRU disturbance observer according to any one of claims 1 to 8, wherein the model predictive control strategy comprises the steps of: Discretizing the unmanned ship linearization model to obtain a discrete state equation; Obtaining expected forward speed through path planning algorithm Desired heading angle Prediction step size Control step size Constructing a prediction model; setting quadratic performance function Minimizing the prediction step size by Tracking error and control amount variation in: Wherein, the In order to predict the error of the output from the desired output, As a matrix of the weights of the errors, In order to control the amount of increase, For controlling the incremental weight matrix; meeting propeller reasoning constraints Is to get future An optimal control quantity sequence at each moment; Selecting a first control amount of the optimal control amount sequence as a target control amount 。
10. The unmanned ship prediction tracking control system based on the GRU disturbance observer is characterized by comprising 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 instruction execution module; the data acquisition module acquires an input vector And transmitting to the GRU disturbance observer module; The GRU disturbance observer module stores a pre-trained GRU disturbance observer that inputs vectors Processing and outputting disturbance estimated value And To a compensation calculation module; The compensation calculation module is used for calculating the disturbance estimated value according to the disturbance estimated value And Calculating the compensation thrust Disturbance estimation value And Instead of actual disturbance values And Obtaining a linearization model of the unmanned ship, wherein the linearization model of the unmanned ship obtains a target control amount through a model predictive control strategy Will compensate for thrust And a target control amount Transmitting to a model predictive control module; the model predictive control module compensates the thrust And a target control amount And transmitting to an output synthesis module; the output synthesis module compensates the thrust And a target control amount Adding to obtain a total target control output The total target control output Transmitting the control instruction to a control instruction execution module; and the control instruction execution module transmits the final total target control output to the unmanned ship propeller for execution.

Description

Unmanned ship prediction tracking control method and system based on GRU disturbance observer Technical Field The application relates to the technical field of marine intelligent equipment control, in particular to an unmanned ship prediction tracking control method and system based on a GRU disturbance observer. Background With the development of ocean economy and intelligent ocean construction, unmanned ships are increasingly widely applied to the fields of ocean mapping, environmental monitoring, water patrol and protection and the like. The precise and stable tracking of the unmanned ship to the expected path or track is one of the core functions. Common tracking control methods in the prior art include PID control, ADRC control, and Model Predictive Control (MPC). Among them, model predictive control is favored because it can explicitly handle system constraints and multi-objective optimization. However, the performance of model predictive control is highly dependent on the accuracy of the predictive model used. The unmanned ship's dynamic model is often complex and is strongly influenced by external environmental factors such as wind, waves, currents, etc. and the hydrodynamic characteristics of the hull itself in actual operation, which can be regarded as a "disturbance" acting on the system. Because of the time-varying nature and uncertainty of these perturbations, it is difficult to build accurate mathematical models, resulting in large errors in model predictive control based on approximation models. In order to solve the problems, the prior art proposes to use a radial basis function neural network (RBF NN) as a disturbance observer, estimate a disturbance value in real time, and compensate a system by using the estimated value, thereby simplifying a dynamics model and improving prediction accuracy of model prediction control. The method effectively improves control performance and reduces manual parameter adjustment. However, the method has the following defects that (1) the radial basis function neural network cannot capture dynamic characteristics of disturbance evolution along with time, such as periodicity of waves and burstiness of gusts, so that estimation accuracy is limited, (2) the motion of an unmanned ship and the disturbance have strong time sequence correlation, disturbance estimation is carried out only by means of state information at the current moment, the value of historical information is ignored, and response to abrupt disturbance is possibly delayed, and (3) the conventional disturbance observer usually only depends on motion and thrust data of the ship body and cannot fully utilize direct environment information provided by external environment sensors such as anemometers and wavemeters. Disclosure of Invention The application provides an unmanned ship predictive tracking control method and system based on a GRU disturbance observer, which can solve the problems of reduced accuracy of an unmanned ship tracking expected path and increased course error caused by low prediction accuracy of model predictive control. In order to achieve the above object, according to a first aspect of the present application, there is provided an unmanned ship predictive tracking control method based on a GRU disturbance observer, specifically comprising the steps of: S1, constructing an unmanned ship kinematic model and an unmanned ship kinematic model; s2, in each control period Acquiring IMU sensor data at the current moment, and respectively transmitting the IMU sensor data to an unmanned ship kinematics model and an unmanned ship dynamics model to respectively obtain the advancing speed at the current momentDifferential forward speedAnd actual disturbance value of unmanned ship propeller moduleAnd; S3, constructing a GRU disturbance observer, and inputting vectorsInput to the GRU disturbance observer for training, the input vectorIncluding forward speedAnd differential forward speedObtaining a pre-trained GRU disturbance observer model after training is completed, and outputting disturbance estimated valuesAnd; S4, according to the disturbance estimated valueAndCalculating the compensation thrustDisturbance estimation valueAndInstead of actual disturbance valuesAndObtaining a linearization model of the unmanned ship, wherein the linearization model of the unmanned ship obtains a target control amount through a model predictive control strategyControlling the target amountAnd compensating for thrustAdding to obtain a total target control output which is finally sent to the unmanned ship propeller module; S5, in the next control periodAnd S2-S4 is repeatedly executed, so that closed-loop control is realized. Further, the step of constructing the unmanned ship kinematic model in the step S1 comprises the following steps: the unmanned ship is positioned as Speed of derivativeRotational speedSpeed of forward travelTransverse velocityAnd heading angleThe unmanned ship kinematics model is construc