A large communication period AUV formation method based on improved distributed model prediction

Abstract

The present invention relates to a long communication cycle AUV formation method based on improved distributed model prediction, which can not only overcome the problems of large calculation amount and communication channel blockage of the traditional pilot-follower method, but also well solve the influence of control delay or failure on the system when using large communication cycle underwater acoustic communication equipment. By using the method proposed in the present invention to update the state vector of the neighboring AUV, the system can effectively overcome the failure of the model predictive control algorithm caused by the long communication interval. In addition, the method proposed in the present invention can be used in the underwater AUV formation, and can also be used in the formation problem of too long communication cycle caused by large-area obstacles on land or water.

Description

Technical Field

[0001] The present invention belongs to the field of automatic control technology, and relates to a formation control method for controlling a multi-underwater vehicle formation. Specifically, in actual control, a state prediction is performed for an AUV formation system with a large communication cycle, so that the estimated state value is closer to the real value, so as to achieve control indicators such as formation obstacle avoidance and formation maintenance. Background Art

[0002] As an underwater operation tool that replaces humans to explore and develop the deep sea, the Autonomous Underwater Vehicle (AUV) has become a research hotspot for researchers from various countries due to its many outstanding advantages. Since there is no need for a tether for signal transmission, the range of AUV activities is greatly expanded, no huge surface support is required, and no human intervention is required after entering the water. AUV can complete many tasks such as seabed detection, autonomous reconnaissance, anti-mine warfare, underwater interference, etc. As the scope of application continues to expand, a single AUV is inefficient or even unable to complete large-scale tasks. Therefore, the collaborative operation of multiple underwater vehicles has become a new form of application and has received more widespread attention.

[0003] In terms of control methods, the existing formation methods can be divided into the leader-follower method, artificial potential field method, virtual structure method, behavior-based method, graph theory-based method and reinforcement learning method. Among them, the leader-follower method, virtual structure method and behavior-based method are used in the majority of formation control. The main goal of cluster formation is to use information such as local relative position or speed between individuals to design feedback control mode and form a stable formation. However, due to the special underwater communication conditions, formation control under communication constraints has become a problem of concern to many scholars. Summary of the invention

[0004] Technical issues to be solved

[0005] Aiming at the problems of low communication frequency and long communication interval caused by hardware problems or communication network congestion in multi-AUV formation, the present invention provides an AUV formation method with a large communication period based on improved distributed model prediction.

[0006] Technical Solution

[0007] A large communication period AUV formation method based on improved distributed model prediction, characterized by the following steps:

[0008] Step 1: In t c moment, use the previous moment t c-1 The calculated optimal control input is used as the initial control input for the next control cycle, and the currently observed state vector is used as the initial state for the next control cycle;

[0009] Step 2: Determine the neighboring AUV through communication or information from the previous control moment;

[0010] Step 3: If the current time is the communication time, then set the current time t c The state vector observation value is packaged with the estimated control sequence of the next control cycle, and information is exchanged with neighboring AUVs; if the current time is not the communication time, proceed directly to step 4;

[0011] Step 4: After obtaining the state information of the neighboring AUV through communication interaction or estimation from the previous control cycle, the state of the neighboring AUV is estimated at each control step, and the state vector of the neighboring AUV is predicted. Then, the state information predicted for the neighboring AUV is substituted into the cost function to solve the optimal control sequence for the corresponding step.

[0012] Step 5: According to the optimal control sequence solved in step 4, take the control amount of the first step as the output and transmit it to the actuator for control, then set c=c+1 and update the control step.

[0013] Further technical solution of the present invention: Step 3 is as follows:

[0014] At the current moment, the system state is sampled through sensors and observers, and the estimated control vector is obtained through equation (3.1)

[0015]

[0016] Among them, T p To control the cycle, Indicates that at t c-1 The optimal control sequence calculated at every moment; Indicates time s = t c-1 +T p The optimal control input is used as the control input of the next control sequence;

[0017] The control vector and the observed state sequence Exchange information with neighboring AUVs.

[0018] A further technical solution of the present invention: The state vector of the neighboring AUV is predicted in step 4 as follows:

[0019] When using the estimated control sequence transmitted by adjacent AUVs, the speed information in the state quantity is added, which is specifically expressed as formula (3-2):

[0020]

[0021] Among them, T p is the control period, t kh is the sampling time in the large cycle communication mode, h is the sampling period, k=1,2,…,N means the current update is the kth sampling period, is the estimated speed of the previous control cycle, which can be used to calculate the missing estimated control vector for the next N steps;

[0022] With the above information of adjacent AUVs, the MPC sub-controller substitutes the relevant information into the optimization function (3.3)

[0023]

[0024] Among them, z -i (s;t c ) is the state value of the adjacent AUV received by the controller. In addition to the above optimization function, the controller calculation process also needs to ensure that the system meets the following conditions and constraints

[0025] z i (s;t c ),z -i (s;t c )∈Z (3.4.1)

[0026] u i (s;t c )∈U (3.4.2)

[0027] z i (t c ;t c )=z i (t c ) (3.4.3)

[0028] z i (t c +T p ;t c )∈Φ i (3.4.4)

[0029]

[0030]

[0031] Among them, s∈[t c ,t c +T p ],z i (s;t c )∈Z and z i (t c +T p ;t c )∈Φi Represent the state constraints and terminal constraints respectively; the cost function in the optimization problem is defined as follows

[0032]

[0033] Among them, F i and Φ i They are

[0034]

[0035] Φ i (z i (t))=γ∑||m(t)|| 2 (3.5.2)

[0036] In the above formula, ||·|| is the norm of any vector in n-dimensional space, A represents the combination of pairs of AUVs in the multi-AUV system, and it is assumed that the AUV system communication network is an undirected graph, that is, if (i, j)∈A, then for AUV i ∈A, then (j, i)∈A and α, β and γ are weighting coefficients, and g(t) and m(t) can be selected according to the needs of the formation process;

[0037] The cost function is specifically expressed as formula (3.6) and formula (3.7)

[0038]

[0039]

[0040]

[0041]

[0042] In the formula, α, β, and γ are weighting coefficients. represents the constraints on the i-th AUV control sequence, μ and ρ are the constraints on the reference distance d i and the adjacent AUV sight angle σ i The weight coefficient Indicates the predicted horizontal coordinate difference between the current AUV and other AUVs in the subsystem, represents the predicted ordinate difference between the current AUV and other AUVs in the subsystem, di is the expected distance between the current AUV and the adjacent AUV, and d1cosθ1, d2cosθ1 represent the expected abscissa differences between the current AUV and AUV2, AUV3, respectively, and d1sinθ1, d2sinθ1 represent the expected ordinate differences between the current AUV and AUV2, AUV3;

[0043] After the above optimization function, the controller will calculate the optimal control sequence within a fixed N control cycles.

[0044] A computer system, characterized in that it comprises: one or more processors, and 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 above-mentioned method.

[0045] A computer-readable storage medium, characterized in that it stores computer-executable instructions, which are used to implement the above method when executed.

[0046] Beneficial Effects

[0047] The present invention provides a long communication cycle AUV formation method based on improved distributed model prediction. This method can not only overcome the problems of large calculation amount and communication channel congestion of the traditional leader-follower method, but also can well solve the influence of control delay or failure on the system when using large communication cycle underwater acoustic communication equipment. By using the method proposed in the present invention to update the state vector of the neighboring AUV, the system can effectively overcome the failure of the model predictive control algorithm caused by the long communication interval. In addition, the method proposed in the present invention can be used in underwater AUV formations, and can also be used in formation problems with long communication cycles caused by large-area obstacles on land or water.

[0048] Compared with the existing formation technology, the AUV formation member state estimation method proposed in the present invention has the following advantages:

[0049] 1. High robustness, strong scalability, and low computational effort

[0050] In the past, most methods adopted for AUV formation control adopted centralized control. Although the centralized control algorithm is simple, the master node has a large amount of calculation and communication volume, and the failure of the master node will cause the overall failure of the formation. The present invention will adopt a distributed model predictive control algorithm and multiple model predictive controllers to avoid the drawbacks of the above centralized control methods, and has higher robustness and better scalability.

[0051] 2. Accurate forecast data

[0052] The acoustic communication equipment used in AUV formation control has a low communication frequency, which causes the input sequence of the controller to be "outdated" and the control effect is difficult to guarantee in terms of real-time formation. By adopting the prediction method proposed in the present invention, the state information of adjacent AUVs can be accurately predicted when the communication cycle is large, thereby meeting the real-time requirements of formation control.

[0053] 3. Wide application areas

[0054] The method of the present invention is not only applicable to the AUV formation problem based on underwater acoustic communication at any depth, but can also be applied to the intelligent agent formation problem under land or air combat conditions where the communication frequency cannot be too high. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0056] Figure 1 Distributed model predictive control architecture;

[0057] Figure 2 The i-th and j-th AUV control vector estimation process;

[0058] Figure 3 The simulation trajectory of obstacle avoidance during the 5s communication cycle;

[0059] Figure 4 Forward distance error of formation in single leader mode;

[0060] Figure 5 Thrust, torque, speed and angular velocity curves during formation process;

[0061] Figure 6 Changes in heading angle during formation process;

[0062] Figure 7 5s communication cycle broken line formation simulation trajectory;

[0063] Figure 8 5s communication cycle zigzag formation forward distance error;

[0064] Fig. 9 5s communication cycle broken line formation thrust, torque, speed and angular velocity curves;

[0065] Fig.10 Discounted formation heading angle curve with 5s communication cycle. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0067] The present invention provides a large communication cycle AUV formation method based on improved distributed model prediction. The method is characterized in that the estimated input value used by each MPC controller is closest to the actual state value at the sampling time by changing the estimation method of the state value of the adjacent AUV. In the ideal distributed MPC controller, it is assumed that the update time of the state information is equal to the information interaction time. However, due to the long communication cycle, the state sequence used by the controller may be "outdated" data from a few seconds or more than ten seconds ago, and the obtained control sequence is naturally not optimal. The present invention modifies the estimation function of the adjacent AUV states and modifies the cost function in the MPC controller at the same time, so that the optimal control sequence can be obtained in the controller itself, and finally the AUV formation is realized.

[0068] The present invention comprises the following steps:

[0069] Step 1: Initialization of the i-th AUV

[0070] After the specific formation task is issued, assuming that the current update time is t c moment, through the previous control moment t c-1 The optimal control vector used is used as the initial control input in the controller at the current moment, and the state value observed by the system at the current moment is used as the initial state of the next control cycle.

[0071] Step 2: Identify Neighboring AUVs

[0072] Before the controller solves the optimization problem, it is necessary to determine the information of the adjacent AUVs at the current moment, including the number of adjacent AUVs, the state of adjacent AUVs, etc. If the current moment is the communication moment, the information of the AUVs in the neighborhood can be directly determined through the communication range. If it is not the communication moment, the number of adjacent AUVs in the previous control cycle and other information are continued.

[0073] Step 3: Interact with neighboring AUVs

[0074] At the current moment, the system state is sampled through sensors and observers, and the estimated control vector is obtained through equation (3.1)

[0075]

[0076] Among them, T p To control the cycle, Indicates that at t c-1 The optimal control sequence calculated at every moment; Indicates time s = t c-1 +T p The optimal control input is used as the control input for the next control sequence.

[0077] The control vector and the observed state sequence Exchange information with neighboring AUVs.

[0078] Step 4: Predict nearby AUVs

[0079] When the communication frequency is sufficient, the state of the adjacent AUV can be predicted in N steps. However, due to the long communication cycle, the information obtained through communication is difficult to meet the data requirements of each control cycle. The method of the present invention proposes to add the speed information in the state when using the estimated control sequence transmitted by the adjacent AUV, which is specifically expressed as formula (3-2)

[0080]

[0081] Among them, T p is the control period, t kh is the sampling time in the large cycle communication mode, h is the sampling period, k=1, 2, ..., N means the current update is the kth sampling period, is the estimated speed of the previous control cycle, which can be used to calculate the estimated control vector missing for the next N steps.

[0082] With the above information of adjacent AUVs, the MPC sub-controller substitutes the relevant information into the optimization function (3.3)

[0083]

[0084] Among them, z -i (s;t c ) is the state value of the adjacent AUV received by the controller. In addition to the above optimization function, the controller calculation process also needs to ensure that the system meets the following conditions and constraints

[0085] z i (s;t c ), z -i (s;t c )∈Z (3.4.1)

[0086] u i (s;t c )∈U (3.4.2)

[0087] z i (t c ;t c )=z i (t c ) (3.4.3)

[0088] z i (t c +Tp ;t c )∈Φ i (3.4.4)

[0089]

[0090]

[0091] Among them, s∈[t c , t c +T p ],z i (s;t c )∈Z and z i (t c +T p ;t c )∈Φ i Represent the state constraints and terminal constraints respectively. In the optimization problem, the cost function is defined as follows

[0092]

[0093] Among them, F i and Φ i They are

[0094]

[0095] Φ i (z i (t))=γ∑||m(t)|| 2 (3.5.2)

[0096] In the above formula, ||·|| is the norm of any vector in n-dimensional space, A represents the combination of pairs of AUVs in the multi-AUV system, and it is assumed that the AUV system communication network is an undirected graph, that is, if (i, j)∈A, then for AUV i ∈A, then (j, i)∈A and α, β and γ are weighting coefficients, and g(t) and m(t) can be selected according to the requirements of the formation process.

[0097] In the present invention, the cost function is specifically expressed as formula (3.6) and formula (3.7)

[0098]

[0099]

[0100]

[0101]

[0102] In the formula, α, β, and γ are weighting coefficients. represents the constraints on the i-th AUV control sequence, μ and ρ are the constraints on the reference distance d i and the adjacent AUV sight angle σ i The weight coefficient Indicates the predicted horizontal coordinate difference between the current AUV and other AUVs in the subsystem, represents the predicted ordinate difference between the current AUV and other AUVs in the subsystem, di is the expected distance between the current AUV and the adjacent AUV, and d1cosθ1, d2cosθ1 represent the expected abscissa difference between the current AUV and AUV2, AUV3, respectively, and d1sinθ1, d2sinθ1 represent the expected ordinate difference between the current AUV and AUV2, AUV3; after the above optimization function, the controller will calculate the optimal control sequence within a fixed N control cycles.

[0103] Step 5: Actuator Control

[0104] The optimal control sequence is obtained by calculating in the previous order Afterwards, it contains control vectors of N steps, and only the optimal control vector of the first step is selected in each control cycle. After the controller outputs to the actuator, it goes to step 1 after one control cycle to update the next optimal control.

[0105] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be included in the protection scope of the present invention.

Claims

1. A large communication cycle AUV formation method based on improved distributed model prediction, characterized in that Here are the steps: Step 1: In t c moment, use the previous moment t c-1 The calculated optimal control input is used as the initial control input for the next control cycle, and the currently observed state vector is used as the initial state for the next control cycle; Step 2: Determine the neighboring AUV through communication or information from the previous control moment; Step 3: If the current time is the communication time, then set the current time t c The state vector observation value is packaged with the estimated control sequence of the next control cycle, and information is exchanged with neighboring AUVs; if the current time is not the communication time, proceed directly to step 4; Step 4: After obtaining the state information of the neighboring AUV through communication interaction or estimation from the previous control cycle, the state of the neighboring AUV is estimated at each control step, and the state vector of the neighboring AUV is predicted. Then, the state information predicted for the neighboring AUV is substituted into the cost function to solve the optimal control sequence for the corresponding step. The prediction of the state vector of the neighboring AUV is specifically as follows: When using the estimated control sequence transmitted by adjacent AUVs, the speed information in the state quantity is added, which is specifically expressed as formula (3.2) Among them, T p is the control period, t kh is the sampling time in the large cycle communication mode, h is the sampling period, k=1,2,…,N means the current update is the kth sampling period, is the estimated speed of the previous control cycle, which can be used to calculate the missing estimated control vector for the next N steps; With the above information of adjacent AUVs, the MPC sub-controller substitutes the relevant information into the optimization function (3.3) Among them, z -i (s;t c ) is the state value of the adjacent AUV received by the controller. In addition to the above optimization function, the controller calculation process also needs to ensure that the system meets the following conditions and constraints z i (s;t c ),z -i (s;t c )∈Z (3.4.1) u i (s;t c )∈U (3.4.2) z i (t c ;t c )=z i (t c ) (3.4.3) z i (t c +T p ;t c )∈Φ i (3.4.4) Among them, s∈[t c ,t c +T p ],z i (s;t c )∈Z and z i (t c +T p ;t c )∈Φ i Represent the state constraints and terminal constraints respectively; the cost function in the optimization problem is defined as follows where F i and Φ i They are Φ i (z i (t))=γ∑||m(t)|| 2 (3.5.2) In the above formula, ||·|| is the norm of any vector in n-dimensional space, A represents the combination of pairs of AUVs in the multi-AUV system, and it is assumed that the AUV system communication network is an undirected graph, that is, if (i, j)∈A, then for AUV i ∈A, then (j,i)∈A and α, β and γ are weighting coefficients, and g(t) and m(t) can be selected according to the needs of the formation process; The cost function is specifically expressed as formula (3.6.1)-(3.6.3) and formula (3.7) In the formula, α, β, and γ are weighting coefficients. represents the constraints on the i-th AUV control sequence, μ and ρ are the constraints on the reference distance d i and the adjacent AUV sight angle σ i The weight coefficient Indicates the predicted horizontal coordinate difference between the current AUV and other AUVs in the subsystem, Indicates the predicted ordinate difference between the current AUV and other AUVs in the subsystem, d i is the expected distance between the current AUV and the adjacent AUV, and d1cosθ1, d2cosθ1 represent the expected horizontal coordinate difference between the current AUV and AUV2, AUV3, respectively, and d1sinθ1, d2sinθ1 represent the expected vertical coordinate difference between the current AUV and AUV2, AUV3; After the above optimization function, the controller will calculate the optimal control sequence within a fixed N control cycles. Step 5: According to the optimal control sequence solved in step 4, take the control amount of the first step as the output and transmit it to the actuator for control, then set c=c+1 and update the control step.

2. According to claim 1, a large communication period AUV formation method based on improved distributed model prediction is characterized in that: Step 3 is as follows: At the current moment, the system state is sampled through sensors and observers, and the estimated control vector is obtained through equation (3.1) Among them, T p To control the cycle, Indicates that at t c-1 The optimal control sequence calculated at every moment; Indicates time s = t c-1 +T p The optimal control input is used as the control input of the next control sequence; The control vector and the observed state sequence Exchange information with neighboring AUVs.

3. A computer system, characterized in that include: One or more processors, and 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 are enabled to implement the method of claim 1.

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

Images