Method for predicting motion position of unmanned ship

Abstract

The invention discloses an unmanned ship motion position prediction method. The method comprises the steps of 1, predicting unmanned ship prediction position information through a kinetic model; 2, obtaining the actual position information of the unmanned ship, and making a difference between the actual position information of the unmanned ship and the predicted position information; 3, taking the state information and the difference value of the unmanned ship as the input and the output of an LSTM neural network respectively, and performing training; 4, obtaining the real initial state of the unmanned ship, and predicting the predicted position information of the unmanned ship in a complete time sequence through the dynamic model; and obtaining the predicted position deviation information of the unmanned ship in the same complete time sequence through an LSTM neural network, and adding the predicted position information of the unmanned ship at each moment and the predicted position deviation information of the unmanned ship to obtain a final position prediction result of the unmanned ship. According to the method, the model prediction error caused by uncertain factors without modeling is compensated, and the problem of poor adaptability to different environments caused by prediction by a single method is avoided.

Description

Technical Field

[0001] The invention relates to the technical field of ship motion position prediction, and in particular to an unmanned ship motion position prediction method based on an IGWO-LSTM hybrid dynamics model. Background Art

[0002] With the continuous development of unmanned technology, unmanned ships play an important role in marine scientific research. They can perform survey, ecological monitoring, water quality monitoring and other tasks efficiently, safely and at low cost. However, unmanned ships often perform navigation tasks in complex and changeable water environments. Due to the size, the interference of uncertain factors such as wind and waves often causes the navigation of unmanned ships to deviate from the expected route, causing sudden and unnecessary troubles to remote control. Therefore, it is very important to predict the position of unmanned ships based on sensor acquisition signals.

[0003] Ship position prediction usually revolves around models that process time series data, where each input channel contains data sampled at fixed time intervals. There are two general approaches to time series prediction models: (1) using a predictor based on a dynamic kinematic model of the ship's forces and velocity; and (2) using a data predictor that learns the ship's behavior based on sampled data.

[0004] First, model-based prediction generally uses various filters to estimate the motion state of the ship or model the ship's motion, thereby obtaining a prediction model of the ship's position in the short term.

[0005] Data-based predictors usually associate a sampled input vector with an output state vector. The process is independent of the physical characteristics of the ship and is generally carried out using support vector machines (SVM), neural networks, AR models, and their variants.

[0006] Traditional prediction methods include model-based prediction and data-based prediction. Model-based prediction predicts the movement of ships based on the ship dynamics modeling equation. However, since the mathematical model cannot accurately describe uncertain factors such as wind and waves, the predicted position of the unmanned ship often deviates from the actual position. Data-driven predictors generally use neural networks and rely on the analysis of data to identify potential patterns and associations. This method cannot explain the reasons and mechanisms behind the movement of ships, but simply makes predictions based on the statistical laws of historical data. Therefore, when faced with new environments or situations, data-driven predictors may not be able to adapt or produce accurate prediction results. Summary of the invention

[0007] In view of the deficiencies in the prior art, the present invention provides a method for predicting the motion position of an unmanned ship to solve the technical problem that the prior art uses a single prediction method and cannot accurately predict the motion position according to actual conditions.

[0008] The present invention provides a method for predicting the motion position of an unmanned ship, comprising the following steps:

[0009] Step 1: Construct a dynamic model of the unmanned ship's navigation. According to the initial state information of the unmanned ship, the unmanned ship's predicted position information within a complete time series is predicted through the dynamic model of the unmanned ship's navigation.

[0010] Step 2: Obtain the actual position information of the unmanned ship in the same complete time series as step 1, and make a difference between the actual position information of the unmanned ship and the predicted position information of the unmanned ship at the corresponding moment;

[0011] Step 3: Using the state information of the unmanned ship as the input of the LSTM neural network, and using the difference obtained in step 2 as the output of the LSTM neural network to train the LSTM neural network;

[0012] Step 4: Obtain the real initial state of the unmanned ship, and predict the predicted position information of the unmanned ship at each moment in a complete time series through the dynamic model of the unmanned ship's navigation; obtain the predicted position deviation information of the unmanned ship at each moment in the same complete time series through the LSTM neural network, and add the predicted position information of the unmanned ship at each moment to the predicted position deviation information of the unmanned ship to obtain the final unmanned ship position prediction result in the complete time series.

[0013] Furthermore, in step 3, the hyperparameters of the LSTM neural network are optimized by the Grey Wolf Algorithm.

[0014] Furthermore, the gray wolf algorithm is an optimized gray wolf algorithm, and the optimization method is: adding a disturbance factor in the search process of α wolf, β wolf and δ wolf, and the specific formula is as follows:

[0015] X 11 =(X α -A1·D α )·α f

[0016] X 22 =(X β -A2·D β )·β1

[0017] X 33 =(X δ -A3·D δ )·α c

[0018] Where, X 11 , X 22 , X 33 are the search activity ranges of α, β, and δ wolves respectively; X α , Xβ , X δ is the position of α, β, δ wolves; D is the distance between the gray wolf and the target; A is the coefficient vector; α f is the disturbance factor for updating the position of α wolf, which makes the exploration range of α wolf transition from large to small; β1 is the disturbance factor for updating the position of β wolf, which is used to control the proportion of the search range of β wolf, and is a random number in [0,1]; α c is the disturbance factor for updating the position of δwolf, which makes δwolf search within the specified range, with the exploration range decreasing linearly, and performing small step search after finding the target.

[0019] Furthermore, the distance D between the gray wolf and the target is expressed as follows:

[0020] D=|C·X p (t)-X(t)|

[0021] Where t is the current iteration number; X p (t) is the current search target position; X(t) represents the current search wolf individual position vector; C is the coefficient vector, and its value range is [0,2].

[0022] Further, the coefficient vector A is expressed as follows:

[0023] A=2ar-a

[0024] Where r is a random vector between [0,1]; a is a coefficient, which decreases linearly from 2 to 0 during the iteration process.

[0025] Furthermore, the disturbance factor α f With α c The formula for obtaining is:

[0026] a f =step·exp(-30t / T)

[0027] a c =step[(Tt) / t]

[0028]

[0029] In the formula, step is the maximum step length of the gray wolf moving within the search activity range; and are the maximum and minimum values ​​of the average values ​​of each latitude in the gray wolf movement space; T is the maximum number of iterations of the gray wolf population; and t is the current number of iterations.

[0030] Furthermore, the disturbance factor β1 is 0.5.

[0031] Furthermore, the position update formula of the ω wolf is:

[0032]

[0033] Where X1, X2, and X3 are the search activity ranges of α wolf, β wolf, and δ wolf without adding disturbance factors:

[0034] X1=X α —A1·D α

[0035] X2=X β —A2·D β

[0036] X3=X δ —A3·D δ

[0037] Beneficial effects of the present invention:

[0038] The present invention uses LSTM long short-term memory neural network prediction to compensate for the prediction error of the unmanned ship motion model caused by unmodeled uncertainties, while avoiding the problem of poor adaptability to different environments when single data-driven prediction is used, which is out of line with the principle of the unmanned ship motion mechanism. It ensures the accuracy of the prediction of the unmanned ship position and track, has a good auxiliary effect on the remote control operation and automatic control of the unmanned ship, improves the efficiency of the unmanned ship in performing navigation tasks, and ensures the safety of personnel and equipment.

[0039] The present invention uses the error between the prediction of the dynamic model and the actual as the supervised training target of the LSTM neural network, and uses the neural network to compensate for the influence of the unmodeled factors in the dynamic model. At the same time, the time series position error is used as the output of the neural network, and is added to the result predicted by the dynamic model to obtain the final position prediction. When the 20s prediction is performed under the tested sea conditions, the position prediction of the single unmanned ship dynamic model is improved by 74%-42%, and the prediction using the same LSTM long short-term memory network is improved by 8%-46%.

[0040] The present invention adopts IGWO algorithm to optimize and find hyperparameter combinations, solving the problem of difficulty in adjusting hyperparameter combinations during training of LSTM neural networks. The optimized Gray Wolf Algorithm used in the present invention is an automated optimization method, which reduces the workload of manual parameter adjustment and can more comprehensively search the hyperparameter space to find a better hyperparameter combination, thereby improving the performance of the neural network model and shortening the parameter adjustment time. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0042] Figure 1is a schematic diagram of a method flow of a specific embodiment of the present invention;

[0043] Figure 2 is a schematic diagram of relative wind direction and angle of attack in a specific embodiment of the present invention;

[0044] Figure 3 is the wind force coefficient of the unmanned ship in a specific embodiment of the present invention;

[0045] Figure 4 is the MPM dual-parameter spectrum in a specific embodiment of the present invention;

[0046] Figure 5 It is the process of solving the response of the ship to the waves;

[0047] Figure 6 It is a schematic diagram of the kinetic model prediction process in a specific embodiment of the present invention;

[0048] Figure 7 This is a schematic diagram of the LSTM neural network cell structure in a specific embodiment of the present invention;

[0049] Figure 8 It is a schematic diagram of the double hidden layer structure of the LSTM neural network in a specific embodiment of the present invention;

[0050] Fig. 9 Schematic diagram of the Grey Wolf Algorithm optimizing LSTM neural network in a specific embodiment of the present invention;

[0051] Fig.10 1 is a schematic diagram of the LSTM neural network training loss rate curve in a specific embodiment of the present invention;

[0052] Fig.11 is the error prediction of the E direction and the N direction under sea state 3 in a specific embodiment of the present invention;

[0053] Fig.12 is the error prediction of the E direction and the N direction under sea state 4 in a specific embodiment of the present invention;

[0054] Fig.13 It is a comparison of the prediction effects of various prediction methods under sea condition 3 in a specific embodiment of the present invention;

[0055] Fig.14 It is a comparison of the prediction effects of various prediction methods under sea condition 4 in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0057] The present invention is further illustrated below in conjunction with specific embodiments. Those skilled in the art should understand that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention, and modifications to various equivalent forms of the present invention fall within the scope defined by the appended claims of this application.

[0058] like Figure 1 As shown, the present invention provides a method for predicting the motion position of an unmanned ship, comprising the following steps:

[0059] Step 1: Construct a dynamic model of the unmanned ship's navigation. According to the initial state information of the unmanned ship, the dynamic model of the unmanned ship's navigation is used to predict the predicted position information of the unmanned ship within a complete time series:

[0060] like Figure 6 As shown, the unmanned ship dynamics model is established as:

[0061]

[0062] In the formula, η=[x n ,y n ,z n ,φ,θ,ψ] T is the vector of position and Euler angle; is the fixed linear speed of the unmanned vessel; is the fixed angular velocity of the unmanned ship; J θ is the motion coordinate system rotation matrix, which can be specifically expressed as:

[0063]

[0064] Among them, s·=sin(·), c·=cos(·), t·=tan(·);

[0065] In the mathematical model of the generalized forces acting on the unmanned ship's hull motion, M, C(v), and D(v) are the rigid body mass matrix, the rigid body Coriolis centripetal force matrix, and the damping matrix, respectively; g(η) is the generalized gravity and buoyancy moment vector; and g0 is the static restoring force and moment vector.

[0066] For an unmanned ship sailing on the water at a certain speed, such as Figure 2 As shown, according to the relative wind speed V rwand the angle of attack γ rw , the forces and moments acting on the hull can be expressed as:

[0067]

[0068] In the formula, and They are the front and side projection areas of the unmanned ship respectively; and are the centroids of the two projected areas above the waterline; L oa is the overall length of the hull; C X , C Y , C Z , C K , C M , C N are the forces and moments of wind on the x, y, and z axes respectively.

[0069] The parameters of the unmanned ship in this embodiment are shown in Table 1 below:

[0070]

[0071] Table 1

[0072] In Table 1, Loa is the total length; B is the ship width; ALw is the transverse projection area (left and right); AFw is the transverse projection area (front and back); A_SS is the lateral projection area; M is the distance from the transverse projection area to the center of mass. The results of calculating the wind force and moment using MATLABR2022b are as follows: Figure 3 shown.

[0073] like Figure 4 As shown in the figure, the Modified Pierson-Moskowitz (MPM) two-parameter spectrum is used and the force response amplitude operator (Force RAO) is introduced to solve the resultant force and moment of the wave disturbance factor. The solution process is as follows: Figure 5 As shown, the spectrum of the ocean wave is expressed as:

[0074] S(w)=Aω -5 exp(-Bω -4 )

[0075] In the formula, H s is the sea state parameter; T Z =0.710T0=0.921T1 is the average zero-crossing wave period; T0 is the wave modal period; T1 is the wave average period.

[0076] Table 2 below shows the significant wave heights and their main ranges under different sea conditions:

[0077]

[0078] Table 2

[0079] The position of the unmanned ship is obtained based on the speed of the unmanned ship. The speed of the unmanned ship system is solved according to the unmanned ship motion dynamics equation. The formula is as follows:

[0080]

[0081] The fourth-order Runge-Kutta method is used to integrate the acceleration of the unmanned ship to obtain the velocity v and position η. The fourth-order Runge-Kutta method is used to solve the differential equation of the unmanned ship system:

[0082] v[k+1]=v[k]+hM —1 [Bu[k]—C(v[k])v[k]—D(v[k])v[k]—g(η[k])]

[0083] η[k+1]=η[k]+hJ θ (η[k])v[k+1]

[0084] The position of the unmanned ship system is obtained by integrating the velocity.

[0085] The x, y, z axis speeds and x, y, z axis speeds of the unmanned ship at the initial moment are obtained through the gyroscope and accelerometer, and the propeller speed and steering gear angle are read through the control board and used as the speed v(k) input at time t0. The speed v(k+1) at the next moment is obtained by the above-mentioned solution of the dynamic model, and the predicted position of the unmanned ship in the complete time series is calculated through cyclic iteration.

[0086] Step 2: Obtain the actual position information of the unmanned ship in the same complete time series as step 1, and make a difference between the actual position information of the unmanned ship and the predicted position information of the unmanned ship at the corresponding moment;

[0087] Step 3: Using the state information of the unmanned ship as the input of the LSTM neural network optimized by the optimized gray wolf algorithm, using the difference obtained in step 2 as the output of the LSTM neural network optimized by the optimized gray wolf algorithm, and training the LSTM neural network optimized by the optimized gray wolf algorithm;

[0088] The optimization process of the gray wolf algorithm is as follows:

[0089] The purpose of improving the gray wolf algorithm is to make the decision weight of α wolf decrease as the target search decision power of β wolf and δ wolf changes. The specific formula for updating the gray wolf search activity range is:

[0090] X 11 =(X α -A1·D α )·α f

[0091] X 22 =(X β -A2·D β )·β1

[0092] X 33 =(X δ -A3·D δ )·α c

[0093] Where, X 11 , X 22 , X 33 Respectively represent the search activity range of α, β, and δ wolves; X α , X β , X δ are the positions of α, β, and δ wolves; D represents the distance between the gray wolf and the target; A is the coefficient vector;

[0094] The distance D is expressed as follows:

[0095] D=|C·X p (t)—X(t)|

[0096] Where t is the current iteration number; X p (t) is the current search target position; X(t) represents the current search wolf individual position vector; C is the coefficient vector, and its value range is [0,2].

[0097] The coefficient vector A is expressed as follows:

[0098] A=2ar-a

[0099] In the formula, r is a random vector between [0,1], and the coefficient a decreases linearly from 2 to 0 during the iteration process. Therefore, the value range of A is [-2,2].

[0100] α f is the disturbance factor for updating the position of wolf α, which makes the exploration range of wolf α transition from large to small; c is the disturbance factor for updating the position of δ wolf, which makes δ wolf search within the specified range, and the exploration range decreases linearly. After finding the target, small step search is performed. f With α c The formula for obtaining is:

[0101] a f =step·exp(-30t / T)

[0102] a c =step[(Tt) / t]

[0103]

[0104] In the formula, step represents the maximum step length of the gray wolf moving within the search activity range, and Respectively represent the maximum and minimum values ​​of the average values ​​of each latitude in the gray wolf movement space. T is the maximum iteration number of the gray wolf population, and t is the current iteration number.

[0105] β1 is the disturbance factor of the β wolf position update. As a proportional coefficient, β1∈(0,1), it can affect the search activity of the β wolf and make the algorithm converge faster. The preferred value is 0.5;

[0106] As the number of algorithm iterations increases, the target search capability of the wolf pack changes, and disturbance factors are added to balance and limit the mutual influence of global and local searches.

[0107] For the acquisition of the position of the ω wolf, in the initial stage, when it is far away from the target, the global search capability and search range should be improved; as the number of iterations increases and it gradually approaches the target, the local search capability should be improved to speed up the search. Therefore, the change process of a can be segmented. When the threshold is set to 0.5, the global and local search capabilities can be better balanced: when a≥0.5, the global search capability is improved; when a<0.5, the local search capability is improved. The update formula of the ω wolf's position is as follows:

[0108]

[0109] Where X1, X2, and X3 are the search activity ranges of α wolf, β wolf, and δ wolf without adding disturbance factors, as follows:

[0110] X1=X α —A1·D α

[0111] X2=X β —A2·D β

[0112] X3=X δ —A3·D δ

[0113] Establish LSTM (Long Short-Term Memory) neural network position error prediction model, the specific structure is divided into: forget gate, input gate, output gate and cell state:

[0114] Forget gate: f t =σ(W f ·[h t—1 , x t ]+b f )

[0115] Input gate: i t =σ(Wi ·[h t—1 ,x t ]+b i ),

[0116] Output gate: o t =σ(W o ·[h t—1 ,x t ]+b o ), h t =o t *tanh(Ct)

[0117] Cell Status:

[0118] The specific structure of the cell Figure 7 shown.

[0119] Use Figure 8 The double hidden layer network structure shown in the figure, according to the network input = [hull three-axis angle, hull three-axis speed, three-axis acceleration, GNSS horizontal position, magnetometer heading, steering gear angle, propulsion motor speed, expected heading], uses IMU (Inertial Measurement Unit) components using gyroscopes, accelerometers and magnetometers to assist GNSS (Global Navigation Satellite System) positioning elements to collect position speed, acceleration and heading, sets the sampling interval to 0.1s, and the continuous sampling time to 60s.

[0120] The error values ​​between the position of the unmanned ship in the NED coordinate plane calculated by the dynamic model and the actual position: ΔN, ΔE and the heading angle collected by the gyroscope are used as the output of the neural network. The network output = [distance error in the N direction, distance error in the E direction, heading angle].

[0121] Considering the limitations of the motion control actuator of the unmanned ship and the working environment shown in Table 3, the navigation process is sampled.

[0122]

[0123] Table 3

[0124] In each case, data was collected from two routes, with a total of 4,800 sets of data, which were divided into training and test sets at a ratio of 80%:20%. In order to speed up the convergence of the model, the collected data were normalized and uniformly planned to the [0,1] interval to ensure the stability of the gradient during the training process.

[0125] like Fig. 9As shown in the figure, the training results of the LSTM neural network are greatly affected by the hyperparameter settings, such as the number of hidden layer neurons, the learning rate setting, and the sliding window size. By mapping the hyperparameters to different wolf species, the gray wolf individuals are driven to find the best position, that is, the optimal hyperparameter combination.

[0126] The optimized LSTM neural network has 128 neurons in hidden layer 1 and hidden layer 2, the dropout of each layer is 0.2, the initial learning rate is set to 0.001, and the learning rate decay method is used:

[0127]

[0128] Where r is the decay rate and n is the number of iterations.

[0129] The sliding window method shown in Table 4 below is adopted, and the window size is set to 3.

[0130]

[0131] Table 4

[0132] In Table 4, w d (t), w a (t) represents environmental influencing factors; i n (t) represents the IMU element measurement data; c n (t) represents the operating data of the power actuator; ΔN(t) and ΔE(t) respectively represent the error prediction in the north-east plane of the NED (North-East-Down) coordinate system. The set hyperparameters are used to train and verify the neural network.

[0133] To verify whether the trained LSTM neural network model is overfitting or underfitting, draw the following Fig.10 The loss curves of the training set and the test set are shown. The root mean square error RMSE (Root Mean Square Error) and the mean absolute percentage error MAPE (Mean Absolute Percent Error) are used as evaluation indicators of the neural network prediction results. The calculation formula is as follows:

[0134]

[0135] In the formula, y i and M represent the predicted value, actual value and the number of test samples respectively.

[0136] The prediction output of the neural network under sea state 1 and sea state 2 is as follows Fig.11 , 12 The error calculation results are shown in Table 5 below:

[0137]

[0138] Table 5

[0139] Step 4: Obtain the real initial state of the unmanned ship, and predict the predicted position information of the unmanned ship at each moment in a complete time series through the dynamic model of the unmanned ship's navigation; obtain the predicted position deviation information of the unmanned ship at each moment in the same complete time series through the LSTM neural network, and add the predicted position information of the unmanned ship at each moment to the predicted position deviation information of the unmanned ship to obtain the final unmanned ship position prediction result in the complete time series.

[0140] First, the unmanned ship dynamics model is iterated several times to output the time series position in the NE plane of the unmanned ship NED coordinate system, and then the time series position is added to the distance error in the N and E directions output by the LSTM neural network to obtain the final predicted time series position of the unmanned ship NED coordinate system. The prediction results of the method of the present invention under two sea conditions are compared with those of the prediction using the dynamics model alone and the prediction using the LSTM neural network alone. Fig.13 , 14 shown.

[0141] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for predicting the motion position of an unmanned ship, characterized in that: The steps include: Step 1: Construct a dynamic model of the unmanned ship's navigation. According to the initial state information of the unmanned ship, the unmanned ship's predicted position information within a complete time series is predicted through the dynamic model of the unmanned ship's navigation. Step 2: Obtain the actual position information of the unmanned ship in the same complete time series as step 1, and make a difference between the actual position information of the unmanned ship and the predicted position information of the unmanned ship at the corresponding moment; Step 3: Using the state information of the unmanned ship as the input of the LSTM neural network, and using the difference obtained in step 2 as the output of the LSTM neural network to train the LSTM neural network; Step 4: Obtain the real initial state of the unmanned ship, and predict the predicted position information of the unmanned ship at each moment in a complete time series through the dynamic model of the unmanned ship's navigation; obtain the predicted position deviation information of the unmanned ship at each moment in the same complete time series through the LSTM neural network, and add the predicted position information of the unmanned ship at each moment to the predicted position deviation information of the unmanned ship to obtain the final unmanned ship position prediction result in the complete time series.

2. The method for predicting the motion position of an unmanned ship according to claim 1, characterized in that: In step 3, the hyperparameters of the LSTM neural network are optimized by the Grey Wolf Algorithm.

3. The method for predicting the motion position of an unmanned ship according to claim 2, characterized in that: The gray wolf algorithm is an optimized gray wolf algorithm. The optimization method is: adding a disturbance factor during the search process of α wolf, β wolf and δ wolf. The specific formula is as follows: X 11 =(X α -A1·D α )·α f X 22 =(X β -A2·D β )·β1 X 33 =(X δ -A3·D δ )·a c Where, X 11 , X 22 , X 33 are the search activity ranges of α, β, and δ wolves respectively; X α , X β , X δ is the position of α, β, δ wolves; D is the distance between the gray wolf and the target; A is the coefficient vector; α f is the disturbance factor for updating the position of α wolf, which makes the exploration range of α wolf transition from large to small; β1 is the disturbance factor for updating the position of β wolf, which is used to control the proportion of the search range of β wolf, and is a random number in [0,1]; α c is the disturbance factor for updating the position of δwolf, which makes δwolf search within the specified range, with the exploration range decreasing linearly, and performing small step search after finding the target.

4. The method for predicting the motion position of an unmanned ship according to claim 3, characterized in that: The distance D between the gray wolf and the target is expressed as follows: D=|C·X p (t)—X(t)| Where t is the current iteration number; X p (t) is the current search target position; X(t) represents the current search wolf individual position vector; C is the coefficient vector, and its value range is [0,2].

5. The method for predicting the motion position of an unmanned ship according to claim 3 or 4, characterized in that: The coefficient vector A is expressed as follows: A=2ar-a Where r is a random vector between [0,1]; a is a coefficient, which decreases linearly from 2 to 0 during the iteration process.

6. The method for predicting the motion position of an unmanned ship according to claim 3, characterized in that: The disturbance factor α f With α c The formula for obtaining is: a f =step·exp(-30t / T) a c =step[(T-t) / t] In the formula, step is the maximum step length of the gray wolf moving within the search activity range; and are the maximum and minimum values ​​of the average values ​​at each latitude in the gray wolf movement space, respectively; T is the maximum number of iterations of the gray wolf population; t is the current iteration number.

7. The method for predicting the motion position of an unmanned ship according to claim 3, characterized in that: The disturbance factor β1 is 0.

5.

8. The method for predicting the motion position of an unmanned ship according to claim 3, characterized in that: The position update formula of the ω wolf is: Where X1, X2, and X3 are the search activity ranges of α wolf, β wolf, and δ wolf without adding disturbance factors: X1=X α —A1·D α X2=X β —A2·D β X3=X δ —A3·D δ 。

Images