Unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning

Abstract

The present application relates to a kind of based on model predictive control and control law learning unmanned ship cluster cooperative tracking control method, belong to unmanned ship formation trajectory tracking control field.The present application is aimed at solving the problems of high computational complexity of existing centralized control algorithm and poor control effect of distributed control algorithm.First, the kinematics and dynamics model of unmanned ship is established, then a distributed dual-mode control algorithm and centralized model predictive control algorithm are designed, and a long short-term memory network (LSTM) is constructed to learn the deep mapping relationship between global control law and distributed control law, then, to solve the problem of poor scalability of offline learning method, an adaptive model switching strategy is designed, which adaptively switches the LSTM model according to the characteristics of the tracking trajectory, reduces the calculation time while improving the cooperative control effect, and ensures that the multi-unmanned ship system can track the time-varying target trajectory in real time with various desired formations.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of unmanned surface vehicle formation trajectory tracking control, and particularly relates to an unmanned surface vehicle swarm cooperative tracking control method based on model predictive control and control law learning. BACKGROUND

[0002] An unmanned surface vehicle (USV) is a water surface carrier platform with environmental perception, intelligent decision-making and autonomous navigation functions. Thanks to its endurance and load capacity, the unmanned surface vehicle plays an important role in water tasks such as environmental monitoring, marine rescue and hydrological exploration. Compared with a single unmanned surface vehicle, an unmanned surface vehicle swarm can more flexibly perform complex water tasks, so research related to unmanned surface vehicle swarms has received more and more attention in recent years. As one of the core problems of unmanned surface vehicle swarm control, the cooperative trajectory tracking task requires the unmanned surface vehicle swarm to maintain a certain formation to track a preset target trajectory, and has a very high research value.

[0003] Thanks to the high control accuracy and rolling optimization of distributed model predictive control (DMPC), the DMPC strategy is often used for trajectory tracking control of unmanned surface vehicle swarms. Compared with centralized MPC, DMPC has multiple sub-controllers, so the computational burden is lower, thereby meeting the real-time requirements of multi-unmanned surface vehicle systems. However, the difference between information sent and received by the unmanned surface vehicle causes information differences in the multi-unmanned surface vehicle system, thereby reducing the control effect of DMPC. In addition, DMPC still needs to solve optimization problems, and its computational complexity cannot meet the high-frequency control requirements of some unmanned surface vehicles. Therefore, how to balance the control effect and computational complexity is a big problem in designing a multi-unmanned surface vehicle cooperative trajectory tracking control algorithm.

[0004] The control law learning strategy based on deep learning can use a deep network to output an optimal control law, avoid solving complex optimization problems, and reduce the calculation time of the control algorithm. Long Short-Term Memory (LSTM) is a kind of time recurrent neural network, which is good at solving sequence to sequence mapping problems (Sequence to Sequence, Seq2Seq). Therefore, the mapping relationship between the distributed control law and the global control law is learned by using an LSTM model, thereby improving the distributed control effect, reducing the solving time of the optimization algorithm, and realizing cooperative trajectory tracking control of the unmanned surface vehicle swarm. SUMMARY

[0005] The technical problem to be solved by the present application is:

[0006] In order to avoid the shortcomings of the prior art, the present application provides a kind of unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning.For the problems of high computational complexity and poor control effect of existing unmanned ship cluster cooperative trajectory tracking control algorithm.

[0007] In order to solve the above technical problems, the technical scheme adopted by the present application is:

[0008] A kind of unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning, characterized by comprising:

[0009] Step 1: establish the trajectory tracking error model of unmanned ship cluster;

[0010] Step 2: using the global information of unmanned ship cluster, based on trajectory tracking error model Design centralized model predictive control algorithm, calculate the optimal global control law;

[0011] Step 3: using the internal interaction information of unmanned ship cluster, based on trajectory tracking error model Design double-mode trajectory tracking control algorithm composed of distributed model predictive control algorithm and feedback control law, calculate the optimal distributed control law;

[0012] Step 4: randomly sample different target trajectories and desired formations, with the distributed control law of the double-mode trajectory tracking control algorithm as input, and with the global control law calculated by the centralized model predictive control algorithm as output, to build a training set for the sequence-to-sequence problem;Based on target trajectory and desired formation Calculate tracking eigenvalue, based on tracking eigenvalue Classification training set;

[0013] Step 5: establish a control law learning strategy based on LSTM model, learn the deep mapping relationship between global control law and distributed control law;Based on the classified training set, train the LSTM model corresponding to the tracking eigenvalue of different categories;

[0014] Step 6: after receiving the target trajectory and desired formation, select the optimal LSTM model according to the tracking eigenvalue;Based on distributed control law Real-time calculation of approximate global control law, based on approximate global control quantity Control unmanned ship cluster to track time-varying trajectory.

[0015] Further technical solutions of the present application: in step 1, establish the trajectory tracking error model of multi-unmanned ship system, specifically:

[0016] First, define the coordinates of unmanned ship i in the world coordinate system as [x s,i ,y s,i ] T , heading angle θ s,i , then, establish the kinematics and dynamics model of unmanned ship

[0017]

[0018] where q s = [x s , y s , θ s ] T is the state of the USV, v x,s , v y,s , w s are the linear and angular velocities of the USV, assuming that all USVs in the multi-USV system have the same model, the velocity update equation of the USV i is:

[0019]

[0020] where m1, m2, m3, d1, d2, d3 are the parameters related to the added mass and damping inside the USV, F x,i is the forward thrust, F w,i is the yawing moment;

[0021] Define the head state of the USV i as q h,i = [x h,i , y h,i , θ i ] T , its head kinematics model is

[0022]

[0023] where h is the length of the midpoint of the USV from the head; when the USV i tracks the target trajectory, the kinematics equation of the virtual tracking target is as follows:

[0024]

[0025] where v , w are the linear and angular velocities of the virtual tracking target; define the tracking error as its update equation is

[0026]

[0027] where the angle difference between the USV i and the virtual tracking target is θ e,i = θ r - θ i ; define the discretized tracking error as:

[0028] q e,i (k+1) = f e (q e,i (k), u e,i (k)) δ + q e,i (k)

[0029] = A(w i (k))q e,i (k)+ Bu e,i (k)

[0030] where, B = delta, delta is control interval,

[0031] The further technical scheme of the present application is that in the step 3, a distributed model predictive control algorithm is designed, and specifically:

[0032] The optimization problem of the distributed model predictive control is shown as follows:

[0033]

[0034] st: q h,i (k|k) = eta h,i (k), q e,i (k+n|k) = f e (q h,i (k+n|k), q r (k+n))

[0035]

[0036] where, q h,i (k+n|k) and q e,i (k+n|k) are the head state of the unmanned ship and the tracking error of the unmanned ship predicted according to the control input u i (k+n|k) and the linear velocity v y,i (k+n|k); and are the head coordinates and tracking errors of the unmanned ship j predicted according to the feasible solution and the linear velocity Q, R and P are weight matrices.

[0037] The further technical scheme of the present application is that in the step 3, a double-mode trajectory tracking control algorithm composed of a distributed model predictive control algorithm and a feedback control law is designed, and specifically:

[0038] In the distributed framework, the optimal control input calculated at the k-1 time and the feedback control law are used to construct the feasible solution as shown below:

[0039]

[0040] where, is the tracking control law, which is defined as:

[0041]

[0042] When the tracking error of the unmanned ship enters the terminal invariant set, the feedback control law is Wherein, the feedback gain K and the design of the terminal invariant set Omega are as follows: for any w i And η e ∈Ω, the inequality (A K (w i )) T PA K (w i )-P≤-Q * Is always true, wherein Q * =Q+K T RK, A K (w i )=A(w i )+BK.

[0043] The further technical scheme of the application is that in step 4, the tracking eigenvalue is calculated based on the target trajectory and the expected formation, and specifically:

[0044]

[0045] Wherein, v r (n) is the speed of the reference trajectory, w r (n) is the angular velocity of the tracking target, d x,i (n) and d y,i (n) are the expected distances between the unmanned ship and the expected tracking target.

[0046] The further technical scheme of the application is that in step 4, the training set is classified based on the tracking eigenvalue, and specifically:

[0047] Randomly select M eigenvalues from the training set as clustering centers, and divide the training set into M classes according to the Euclidean distance between the remaining eigenvalues and the clustering centers;

[0048] Recalculate the center point of each cluster Wherein N D,j Is the number of data set D j Of the jth class, and c j Is the center point of the jth class;

[0049] Continue to classify according to the new center point until the clustering center no longer changes obviously.

[0050] The further technical scheme of the application is that in step 5, the control law learning strategy based on the LSTM model is established to learn the deep mapping relationship between the global control law and the distributed control law, and specifically:

[0051] In the training set, the global control law is defined as The distributed control law is The tracking error predicted according to the distributed control law is The tracking error of the unmanned ship j predicted according to the cluster information received by the unmanned ship i and the distributed control law is

[0052] The mapping relationship between the tracking error and the distributed control law and the global control law is learned through an encoder-decoder model based on an LSTM network:

[0053]

[0054] Wherein, the output of the LSTM model is the deviation between the global control law and the distributed control law The mapping relationship between the distributed control law and the global control law is found by minimizing the mean square error (MSE) loss function.

[0055] The further technical solution of the application is: in step 6, after receiving the target trajectory and the expected formation, the optimal LSTM model is selected according to the tracking eigenvalue; the approximate global control law is calculated in real time based on the distributed control law, and the unmanned ship cluster is controlled to track the time-varying trajectory based on the approximate global control quantity, specifically:

[0056] The unmanned ship i extracts a segment of the target trajectory with a length of N, and the tracking eigenvalue β of the trajectory is L,i The optimal LSTM model is selected;

[0057] The double-mode trajectory tracking control algorithm is used to output the distributed control law in real time And predict the state and control input of other unmanned ships;

[0058] The difference between the global control law and the distributed control law is estimated by using the LSTM model So as to calculate the approximate global control law in real time

[0059] Finally, the first control input in the approximate global control law is The forward thrust F of the unmanned ship is calculated x,i And the yawing moment F w,i , realizing the trajectory tracking control of the unmanned ship i, thereby completing the cooperative trajectory tracking control of the unmanned ship cluster.

[0060] A computer system, characterized in that it comprises: one or more processors, a computer readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors realize the above-mentioned method.

[0061] A computer readable storage medium storing computer executable instructions, which when executed, implement the method described above.

[0062] The present application has the following beneficial effects:

[0063] The present application provides a method for tracking control of unmanned ship cluster based on model predictive control and control law learning, which first designs a feedback control law on the basis of DMPC algorithm, constructs a double-mode trajectory tracking control framework, and designs an LSTM model to learn the deep mapping relationship between the global control law calculated by centralized MPC and the distributed control law, finally forms a control law learning strategy based on LSTM model and model predictive control. The present application uses LSTM model and feedback control law to obtain approximate global control law, without directly calculating the rolling optimization problem, solves the problem of high computational complexity of traditional MPC algorithm, and improves the control effect of distributed control algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0064] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:

[0065] Figure 1 The flowchart of the present application.

[0066] Figure 2 The schematic diagram of trajectory tracking of unmanned ship cluster.

[0067] Figure 3 The schematic diagram of time-varying trajectory tracking of unmanned ship cluster.

[0068] Figure 4 The schematic diagram of time-varying trajectory tracking of unmanned ship cluster with time-varying formation. DETAILED DESCRIPTION

[0069] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0070] It is to be understood that the terms "first", "second", and the like, used in the description and the claims of the application, as well as the above-described drawings, are used to distinguish between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of accomplishing the same objectives stated in the claims using other embodiments or approaches. Moreover, the terms "comprise", "have", and any variations thereof, are intended to cover a non-exclusive inclusion, such that processes, methods, articles, or apparatuses that comprise, have, or are otherwise including a list of steps or elements do not necessarily comprise, have, or are otherwise including only those steps or elements but can include other steps or elements not expressly listed or inherent to such processes, methods, articles, or apparatuses.

[0071] For those skilled in the art to better understand the present application, the present application will be described in detail below in conjunction with specific embodiments.

[0072] Embodiment 1:

[0073] The unmanned ship cluster cooperates to track a preset target trajectory in a fixed formation, including the following steps:

[0074] Step 1: Establish a trajectory tracking error model of the multi-unmanned ship system:

[0075] First, define the coordinates of the unmanned ship i in the world coordinate system as [x s,i ,y s,i ] T , the heading angle as θ s,i , and then establish the kinematics and dynamics model of the unmanned ship

[0076]

[0077] Where q s,i =[x s,i ,y s,i ,θ s,i ] T is the state of the unmanned ship, v x,i , v y,i , w i is the linear and angular velocity of the unmanned ship, and assuming that the models of all unmanned ships inside the multi-unmanned ship system are the same, the velocity update equation of the unmanned ship i is:

[0078]

[0079] Where m1, m2, m3, d1, d2, d3 are parameters related to the added mass and damping inside the unmanned ship, F x,i is the forward thrust, F w,i is the yawing moment.

[0080] Define the bow state of unmanned vessel i as q h,i =[x h,i ,y h,i ,θ i ] T The bow kinematics model is then...

[0081]

[0082] Where h is the distance from the midpoint of the unmanned surface vessel to the bow, and the control input u i =[v x,i ,w i ] T When the unmanned vessel i tracks the target trajectory, the kinematic equations of the virtual target are as follows:

[0083]

[0084] in, Let the linear velocity and angular velocity of the virtual target be denoted as . Define the tracking error as . Its update equation is

[0085]

[0086] in, The angle difference between the unmanned vessel i and the virtual tracking target is θ e,i =θ r -θ i To configure the algorithm in an embedded system, the discretized tracking error is defined as...

[0087]

[0088] in, B = δ, where δ represents control.

[0089] Step 2: Utilize the global information of the multi-unmanned vessel system to design an unmanned vessel swarm cooperative trajectory tracking algorithm based on centralized model predictive control, and calculate the optimal cooperative control law for the swarm.

[0090] During the offline phase, a centralized cost function is constructed based on global information within the multi-unmanned surface vessel system:

[0091]

[0092] in, It refers to the number of unmanned vessels within the unmanned vessel cluster. It is a collection of unmanned vessels excluding unmanned vessel i. and They are based on control input and linear velocity The predicted heading states and tracking errors of USVs. Meanwhile, the tracking errors of all USVs are updated. and are the tracking error of USV j and the control input constraint, Q, R, Q c and P are weight matrices, and N is the prediction horizon of model predictive control. is the terminal invariant set, which indicates that the tracking errors of all USVs will enter the terminal invariant set at the end of the prediction horizon.

[0093] Step 3: Design a dual-mode trajectory tracking control framework composed of distributed model predictive control algorithm and feedback control law using the interaction information between USVs;

[0094] When calculating online, due to the influence of internal information transmission time of multi-USV system, USVs cannot obtain the states of other USVs in the cluster in real time. In order to make full use of the computing power of each USV, the optimization problem of distributed model predictive control is as follows:

[0095]

[0096] where q h,i (k+n|k)and q e,i (k+n|k)are the predicted heading states and tracking errors of USVs according to the control input u i (k+n|k)and the linear velocity v y,i (k+n|k). and are the predicted heading coordinates and tracking errors of USV j according to the feasible solution and the linear velocity Unlike centralized MPC, the states of other USVs in distributed MPC are predicted according to the cluster information received by USV i at the current time.

[0097] Step 4: Randomly sample different trajectories and desired formations, and take the distributed control law calculated by the dual-mode trajectory tracking control algorithm as input, and take the global control law calculated by centralized model predictive control as output, to construct a training set of sequence-to-sequence (Seq2Seq) problem.

[0098] Randomly sample the target trajectory, and define the distributed control law calculated by the dual-mode trajectory tracking control framework as is the tracking error predicted according to the distributed control law, is the tracking error of USV j predicted according to the cluster information received by USV i and the distributed control law. For the same trajectory, the optimal global control law calculated by centralized MPC and the cluster information is After that, take the distributed control law The predicted tracking error And the tracking error of other unmanned ships The input is the global control law The training set is constructed for the output.

[0099] Step 5: Establish a control law learning strategy based on an LSTM model, learn the deep mapping relationship between the global control law and the distributed control law, and design an adaptive model switching strategy for the time-varying target trajectory;

[0100] Step 5.1: Design an encoder-decoder model structure based on an LSTM network to solve the Seq2Seq problem in step 4;

[0101] In the distributed framework, the optimal control input u i * (k+n|k-1) and the feedback control law are constructed as shown below.

[0102]

[0103] wherein, The tracking control law is defined as:

[0104]

[0105] When the unmanned ship tracking error enters the terminal invariant set, the feedback control law is wherein, the feedback gain K and the terminal invariant set Ω are designed as shown below: for any w i and η e ∈Ω, the inequality (A K (w i )) T PA K (w i )-P≤-Q * is always true, wherein Q * =Q+K T RK, A K (w i )=A(w i )+BK.

[0106] Then, the mapping relationship between the tracking error and the distributed control law and the global control law is learned through an encoder-decoder model based on an LSTM network.

[0107]

[0108] wherein, the output of the LSTM model is the deviation between the global control law and the distributed control law The mapping between the distributed control law and the global control law is found by minimizing the mean square error (MSE) loss function.

[0109] Step 5.2: Analyze the unmanned ship model in step 1, design the tracking characteristic value related to the target trajectory and the expected formation;

[0110] For the tracking problem of time-varying target trajectory, an adaptive model switching mechanism is designed, and a segment of sub-target trajectory η 1→N,r (k) is extracted, whose length is equal to the prediction horizon of model predictive control, then the most suitable control strategy is selected according to the maximum speed of the sub-target trajectory and the expected formation, and the tracking characteristic value is defined as

[0111]

[0112] Where v r (n) is the speed of the reference trajectory, w r (n) is the angular velocity of the tracking target, d x,i (n), d y,i (n) are the x, y distances between the unmanned ship and the expected tracking target.

[0113] Step 5.3: Through the clustering algorithm, the data set in step 4 is classified according to the tracking characteristic value, and then the LSTM model of different categories is trained respectively, and the adaptive LSTM model is selected according to the tracking characteristic value;

[0114] First, M cluster centers are randomly selected from the data, and the training set is divided into M categories according to the Euclidean distance. Then, the center point of each cluster is recalculated Where N D,j is the number of data sets D j of the jth category, and c j is the center point of category j. Finally, according to the new center point, continue to divide and classify, and repeat the above steps until the cluster center no longer changes significantly.

[0115] Step 6: After receiving the target trajectory and the expected formation, the optimal LSTM model is selected according to step 5, and the approximate global control law is calculated in real time based on the distributed control law in step 4, to control the unmanned ship cluster to track the time-varying trajectory;

[0116] First, the unmanned ship i extracts a segment of sub-target trajectory with length N, calculates its tracking characteristic value β L,i , and selects the optimal LSTM model accordingly. Then, using the dual-mode control framework, the distributed control law is output in real time, and the states and control inputs of other unmanned ships are predicted according to the received cluster information. After that, the LSTM model is used to estimate the difference between the global control law and the distributed control law Thus, the approximate global control law is calculated in real time Finally, the first control input in the approximate global control law The forward thrust F of the unmanned ship is calculated x,i And the yawing moment F w,i The trajectory tracking control of the unmanned ship i is realized, so that the cooperative trajectory tracking control of the unmanned ship cluster is completed.

[0117] Embodiment 2: Unmanned ship cluster tracking time-varying target trajectory

[0118] As Figure 3 shown, when the unmanned ship cluster is controlled to track the time-varying trajectory, the conventional trajectory tracking algorithm usually relies on the maximum value of the target trajectory speed to design a conservative feedback control law, which reduces the control effect of the trajectory tracking algorithm. Therefore, the unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning designed by the present application contains different target trajectories in the training set during offline training, then, consistent with the calculation steps in embodiment 1, the training set is classified by extracting the tracking features of the target trajectory, and then the LSTM model suitable for different tracking target speeds is trained. During online tracking, the tracking target is classified by extracting the features of the tracking trajectory in real time, so as to select the optimal LSTM model and adaptively output the approximate global control input, which improves the control effect of the distributed model predictive control algorithm while reducing the optimization complexity and conservativeness of the tracking control algorithm.

[0119] Embodiment 3: Unmanned ship cluster tracking time-varying target trajectory with time-varying formation

[0120] As Figure 4 shown, when the unmanned ship cluster is controlled to track the time-varying trajectory, the conventional trajectory tracking algorithm usually relies on the maximum value of the target trajectory speed and the farthest distance between the unmanned ship and the tracking target in the time-varying formation to design a conservative feedback control law, which reduces the control effect of the trajectory tracking algorithm. Therefore, the unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning designed by the present application contains different target trajectories and different cooperative formations in the training set during offline training, then, consistent with the calculation steps in embodiment 1, the training set is classified by extracting the tracking features of the target trajectory and the expected formation, and then the LSTM model suitable for different tracking target speeds and formations is trained.

[0121] During online tracking, the features of the tracking trajectory and the expected formation are extracted in real time, and the tracking trajectory and the expected formation are classified, so as to select the optimal LSTM model. In the design of the dual-mode control framework based on distributed MPC and feedback control law, the feedback control law requires that the tracking error be in the terminal invariant set. When the mutation of the new expected formation exceeds the terminal invariant set, that is The feedback control law cannot output a feasible solution. Therefore, the application designs a segmented formation transformation strategy, assuming that the desired formation does not change at k-1, at which time the tracking error is in the terminal invariant set, i.e. q e,i (k-1)∈Ω, then at k, the formation transformation needs to ensure that the tracking error is located in the terminal invariant set, and the tracking target state is defined as η r (k)=f r (η r (k-1),u r (k-1))+Δd, wherein the transformation distance Δd of the formation needs to satisfy Δd∈{d e |f e (q h,i (k),q r (k)+d e )∈Ω}. Then, the real-time extracted target trajectory and the tracking features of the desired formation are selected according to the trajectory features to select the optimal LSTM model. Finally, the calculation of the approximate global control input in real-time example 1 is consistent, so as to control the unmanned ship cluster to track the time-varying trajectory in the time-varying formation.

[0122] The above merely describes the specific embodiments of the application, but the protection scope of the application is not limited thereto, and any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the application, and these modifications or replacements should be covered within the protection scope of the application.

Claims

1. An unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning, characterized in that, The method comprises the following steps: Step 1: establishing a trajectory tracking error model of the unmanned ship cluster; In step 1, the trajectory tracking error model of the multi-unmanned ship system is established, and specifically: First, define the unmanned ship The coordinates in the world coordinate system are The heading angle is Subsequently, establish the kinematics and dynamics model of the unmanned ship wherein, is the state of the USV, is the linear and angular velocity of the USV, assuming that all USVs inside the multi-USV system have the same model, then the velocity update equation of the USV is wherein, , , , , , is an internal parameter of the unmanned ship related to added mass and damping, is a forward thrust, is a yaw moment; Definition of unmanned ship The bow state of the ship is The kinematic model of the bow of the ship is wherein, L is the length of the unmanned ship from the midpoint to the bow; when the unmanned ship When tracking the target trajectory, the kinematic equation of the virtual tracking target is as follows: where, is the linear and angular velocity of the virtual tracking target; the tracking error is defined as The update equation is wherein, unmanned ship the angle difference between the virtual tracking target and the target is the discretized tracking error is defined as: wherein , , is a control interval, ; Step 2: using the global information of the unmanned ship cluster, designing a centralized model predictive control algorithm based on the trajectory tracking error model to calculate the optimal global control law; Step 3: using the internal interaction information of the unmanned ship cluster, designing a dual-mode trajectory tracking control algorithm composed of a distributed model predictive control algorithm and a feedback control law based on the trajectory tracking error model to calculate the optimal distributed control law; in step 3, the distributed model predictive control algorithm is designed, and specifically: The optimization problem of the distributed model predictive control is as follows: wherein, and are the predicted unmanned ship heading state and the unmanned ship tracking error according to the control input and the linear velocity ; and are the ship heading coordinates and the tracking error of the unmanned ship predicted according to the feasible solution and the linear velocity , , , , is a weight matrix; In step 3, the dual-mode trajectory tracking control algorithm composed of the distributed model predictive control algorithm and the feedback control law is designed, and specifically: In a distributed framework, the optimal control input is computed at each time instant using the optimal control input and the feedback control law builds a feasible solution as shown below: wherein , is the tracking control law defined as: When the tracking error of the unmanned ship enters the terminal invariant set, the feedback control law is where the feedback gain and the terminal invariant set are designed as follows: for any and , the inequality is always true, where , ; Step 4: randomly sampling different target trajectories and desired formations, taking the distributed control law of the dual-mode trajectory tracking control algorithm as the input and taking the global control law calculated by the centralized model predictive control algorithm as the output to construct a training set of a sequence-to-sequence problem; calculating a tracking feature value based on the target trajectory and the desired formation, and classifying the training set based on the tracking feature value; Step 5: establishing a control law learning strategy based on an LSTM model to learn the deep mapping relationship between the global control law and the distributed control law; training the LSTM model corresponding to the tracking feature value of different categories based on the classified training set; Step 6: after receiving the target trajectory and the desired formation, selecting the optimal LSTM model according to the tracking feature value; calculating an approximate global control law based on the distributed control law in real time, and controlling the unmanned ship cluster to track the time-varying trajectory based on the approximate global control law.

2. The method of claim 1, wherein, In step 4, the tracking feature value is calculated based on the target trajectory and the desired formation, and specifically: wherein, is the velocity of the reference trajectory, is the angular velocity of the tracking target, , is the desired distance of the unmanned ship from the desired tracking target.

3. The method of claim 2, wherein, In step 4, the training set is classified based on the tracking feature value, and specifically: randomly selected from the training set a feature value as a cluster center, and divide the training set into classes according to the Euclidean distance between the remaining feature values and the cluster center; Recalculate the centroid of each cluster. ,in It is the first Datasets of each category The number of It is a category The center point; The new center point is continuously classified until the clustering center no longer changes obviously.

4. The method of claim 3, wherein, In step 5, the control law learning strategy based on the LSTM model is established to learn the deep mapping relationship between the global control law and the distributed control law, and specifically: In the training set, the global control law is defined as , and the distributed control law is defined as , is the tracking error predicted according to the distributed control law, is the tracking error of the unmanned ship predicted according to the cluster information received by the unmanned ship and the distributed control law. The mapping relationship between the tracking error and the distributed control law and the global control law is learned through an encoder-decoder model based on the LSTM network: Wherein, the output of the LSTM model is the deviation between the global control law and the distributed control law ; By minimizing the mean square error (MSE) loss function, the mapping relationship between the distributed control law and the global control law is found.

5. The method of claim 4, wherein, In step 6, after receiving the target trajectory and the desired formation, the optimal LSTM model is selected according to the tracking feature value; an approximate global control law is calculated in real time based on the distributed control law, and the unmanned ship cluster is controlled to track the time-varying trajectory based on the approximate global control law, and specifically: Unmanned ship i extracting a segment of a sub-target trajectory with a length of a tracking feature value based on the trajectory selecting an optimal LSTM model; The double-mode trajectory tracking control algorithm is used to output a distributed control law in real time and predict the states and control inputs of other unmanned ships Estimating the difference between global control law and distributed control law using LSTM model Thus, the approximate global control law is calculated in real time ; Finally, according to the first control input in the approximate global control law calculating the forward thrust of the unmanned ship and the yaw moment , realizing the trajectory tracking control of the unmanned ship , thereby completing the cooperative trajectory tracking control of the unmanned ship cluster.

6. A computer system, characterized by The method comprises the following steps: One or more processors, a computer readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method of claim 1.

7. A computer-readable storage medium, characterized in that Computer executable instructions are stored, and the instructions are used to implement the method of claim 1 when executed.

Images