A method for UUV encirclement based on distributed model predictive control

Abstract

The present application relates to the technical field of automatic control, in particular to a UUV hunting method based on distributed model predictive control, which comprises establishing a distributed control system structure, obtaining a prediction state sequence of a current UUV and an optimal prediction state sequence of a neighborhood UUV through an outer controller, performing optimization solution to obtain an optimal prediction state sequence and an optimal control sequence of the current UUV respectively, and adding a current UUV control compensation to the optimal prediction state sequence and the optimal control sequence of the current UUV to obtain a final control quantity of the current UUV. The present application is responsible for communication interaction, target detection and optimal control sequence calculation through the outer controller, and the inner controller is responsible for compensating for the uncertain part of the system model and resisting external interference.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic control, and particularly relates to a UUV hunting method based on distributed model predictive control and a system thereof. BACKGROUND

[0002] An unmanned undersea vehicle (UUV) is a kind of underwater carrier that uses various sensors and other task modules to perform tasks such as mine sweeping, reconnaissance, surveying and intelligence gathering, and performs autonomous navigation, autonomous decision-making and autonomous control through its own energy equipment, navigation equipment and execution mechanism.

[0003] With the increase of marine task demand, a single vehicle is limited by poor redundancy, small task range and low work efficiency. Using a multi-UUV system to perform common tasks can well overcome the above problems. Multi-UUV cooperation can increase the redundancy and robustness of the system, improve the efficiency of task execution, and reduce the cost, so more and more scientific research institutions and scholars pay attention to the multi-UUV cooperation system. The multi-UUV system has the characteristics of spatial distribution, time distribution, information distribution, resource distribution and functional distribution.

[0004] As a new type of cooperation platform, the multi-UUV cooperative formation system not only enriches and expands the intelligent technology system of the sea unmanned, but also is an important member of the "sea, land and air" multi-body cooperative operation system. The multi-UUV cooperative formation system breaks through the limitations of single-UUV execution of operation tasks, provides more possibilities for promoting marine development and exploration, at the same time, the development and batch equipment of the multi-UUV system on the sea have important significance for carrying out deep-sea scientific research, promoting the construction of "transparent ocean" engineering, and maintaining marine safety.

[0005] Traditional multi-agent control methods mostly use centralized model control algorithm, the controller design is more convenient, and the internal communication scheme of the system is relatively simple, but in a large-scale agent group, the application of centralized MPC controller is easy to appear the problems of large amount of calculation, large communication load, low fault tolerance rate and reliability; in view of the complexity and difficulty of non-global communication in the multi-UUV system, the local communication mode is improved, and the distributed model prediction can well solve the above problems, and has better processing effect for the multi-input multi-output system, which improves the robustness of the system and significantly reduces the amount of calculation and communication. SUMMARY

[0006] In order to solve the above technical problems, the present application mainly aims at the complexity and difficulty of non-global communication of the multi-UUV system, improves the local communication mode, and proposes a UUV encirclement method based on distributed model predictive control. The method designs a double-layer MPC control framework for the multi-UUV encirclement control task. The outer controller is responsible for communication interaction, target detection, and calculation of the optimal control sequence, and the inner controller is responsible for compensating for the uncertain part of the system model and resisting external interference. The communication mode in the group adopts the local communication mode, and the two-dimensional grid coding method is used to limit the neighborhood of the individual, so that the communication neighbors of each UUV are simplified to two, which is beneficial to reduce the communication pressure and the computational complexity in the large multi-UUV system.

[0007] The first object of the present application is to provide a UUV encirclement method based on distributed model predictive control, which is used for controlling multiple UUVs to encircle a moving target, comprising:

[0008] A distributed control system structure is established, including an inner controller and an outer controller corresponding to each UUV;

[0009] The outer controller is used to obtain the predicted state sequence of the current UUV and the optimal predicted state sequence of the neighborhood UUV, and the optimal predicted state sequence and the optimal control sequence of the current UUV are obtained by optimization solution;

[0010] The inner controller is used to subtract the predicted state sequence of the current UUV from the actual measurement system state to obtain an error, and a error model is defined by the error, and the control compensation amount of the current UUV is obtained by solving the error model;

[0011] The control compensation amount of the current UUV is added to the optimal predicted state sequence and the optimal control sequence of the current UUV to obtain the final control amount of the current UUV;

[0012] After a control period, the outer controller and the inner controller are used to obtain the final control amount of the next period, and the current UUV is controlled by the encirclement control.

[0013] Preferably, the outer controller corresponding to each UUV is responsible for target detection, information interaction between neighborhoods, and solution of the optimal predicted state and the optimal control sequence; and the inner controller corresponding to each UUV takes the optimal predicted state output by the outer controller as a reference state to compensate for the error of the current UUV and obtain the control compensation amount.

[0014] Preferably, it further comprises:

[0015] A group local communication network is established for local information interaction between UUVs;

[0016] The optimal predicted state sequence and optimal control sequence of the current UUV are obtained by the outer controller and uploaded to the local communication network for use by the outer controller of the neighboring UUV.

[0017] Preferably, the neighborhood UUV set corresponding to each UUV is obtained by applying two-dimensional grid coding processing in the local communication network to obtain the neighborhood UUV set of the current UUV.

[0018] Preferably, the predicted state sequence of the current UUV is obtained according to the following steps:

[0019] Constructing nonlinear UUV dynamic model equations;

[0020] The current state of the UUV is predicted based on the nonlinear UUV dynamic model equations to obtain the predicted state sequence of the current UUV.

[0021] Preferably, during the optimization process, each UUV maintains the following three control objectives:

[0022] (1) Using the moving target as the center, converge the encirclement radius to a fixed expected value;

[0023] (2) The UUV surrounds the target with the desired angular velocity;

[0024] (3) The UUVs distributed around the moving target have the same angular intervals in pairs;

[0025] Among them, the moving target is the target to be captured.

[0026] Preferably, after one control cycle, the UUV obtains the final control quantity for the next cycle again through the outer and inner controllers, including:

[0027] Currently, after one control cycle, the UUV contains control vectors of N steps. In each control cycle, only the optimal control vector of the first step is selected. After the controller outputs to the actuator, the final control quantity for the next cycle is obtained again through the outer controller and the inner controller after one control cycle.

[0028] Preferably, the control compensation amount of the current UUV is obtained by solving the error model, including:

[0029] The feedback gain K of the UUV system in steady state is selected as the solution of the linear quadratic regulator of the error model, and is used as the control compensation quantity of the current UUV.

[0030] Preferably, the outer controller is a DMPC-based containment controller; the inner controller is a Tube-based MPC controller.

[0031] The second object of the application is to provide a UUV hunting system based on distributed model predictive control, comprising:

[0032] A control system construction module is configured to establish a distributed control system structure, including an inner controller and an outer controller corresponding to each UUV.

[0033] A control sequence calculation module is configured to obtain a predicted state sequence of a current UUV and an optimal predicted state sequence of a neighboring UUV through the outer controller, and to obtain the optimal predicted state sequence and the optimal control sequence of the current UUV through optimization.

[0034] A control compensation module is configured to obtain an error by subtracting a predicted state sequence of a current UUV from a state quantity of an actual measurement system through the inner controller, and to obtain a control compensation quantity of the current UUV by defining and solving an error model based on the error.

[0035] A control module is configured to obtain a final control quantity of the current UUV by adding the control compensation quantity of the current UUV to the optimal predicted state sequence and the optimal control sequence of the current UUV.

[0036] The application has at least the following beneficial effects:

[0037] The application provides a UUV hunting method based on distributed model predictive control.

[0038] The application improves the robustness and stability of the system through local communication.

[0039] The application designs a two-dimensional grid coding communication method to reduce the node calculation complexity and communication pressure.

[0040] The method provided by the application can realize local information interaction between UUVs, and the state sequence of the neighborhood UUV and the predicted control sequence can be obtained through local communication. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 A multi-UUV hunting control system is provided.

[0042] Figure 2 A schematic diagram of a distributed control structure is provided.

[0043] Figure 3 A schematic diagram of a two-dimensional grid mapping in a group is provided.

[0044] Figure 4 A trajectory of the multi-UUV surrounding a dynamic target is provided.

[0045] Figure 5 A distance error in the hunting process is provided.

[0046] Figure 6 A separation angle of each UUV in the hunting process is provided.

[0047] Figure 7 A hunting angular velocity change is provided. DETAILED DESCRIPTION

[0048] In order to illustrate the technical means and effects adopted by the application to achieve the predetermined application purposes, the following will be described in detail in combination with embodiments.

[0049] The overall control system designed by the application is divided into two parts, as shown in Figure 1 The outer controller is responsible for target detection, information interaction between neighborhoods, and solving of the optimal predicted state and the optimal control sequence, and the inner controller takes the optimal predicted state output by the outer controller as a reference state to perform error compensation on the current underactuated UUV. Both the two controllers are designed based on the model predictive control algorithm.

[0050] Referring to Figure 1 Based on the multi-UUV hunting control system, each UUV applies two-dimensional grid coding processing in the local communication topology to obtain a neighborhood set of the ith unmanned ship, so as to exchange target information and state information. In the distributed MPC controller, the hunting control rule and the control target are used for optimization calculation to obtain the optimal control sequence and the predicted state sequence. The inner controller obtains the error by subtracting the state quantity of the actual measurement system from the predicted state sequence, and then defines an error model through the error to design a control law, so as to obtain the control compensation quantity and add the optimal control sequence obtained by the outer DMPC controller to output the final control quantity.

[0051] Distributed controller is more convenient to solve the control problem of multi-agent system, so the distributed model predictive control algorithm is designed to improve the reliability of the system.

[0052] In the following examples, five UUVs and one dynamic target are taken as examples, the controller described above is applied to simulate and verify the multi-UUV hunting task, and the results are shown, and then the change curves of each state quantity designed in the hunting process are shown and analyzed.

[0053] In the following examples, the outer controller is a DMPC-based hunting controller, and the inner controller is a Tube-based MPC controller.

[0054] The application provides a UUV hunting method based on distributed model predictive control, which is used for controlling multiple UUVs to hunt a dynamic target, and comprises the following steps:

[0055] Step 1, establishing a distributed control structure.

[0056] The distributed control system structure is established, which comprises an inner controller and an outer controller corresponding to each UUV;

[0057] The outer controller corresponding to each UUV is responsible for target detection, information interaction between adjacent areas, and solving of optimal predicted states and optimal control sequences; and the inner controller corresponding to each UUV takes the optimal predicted state output by the outer controller as a reference state to perform error compensation on the current UUV to obtain a control compensation amount.

[0058] The nonlinear state equation of the whole system can be rewritten in the following form:

[0059] f(x(k),u(k))=[f1f2…f N ]

[0060] f i (x i (k),u i (k)), (i=1,2,…,N)

[0061] Wherein, x(k) is the state of the UUV at the k moment, and u(k) represents the control input at the k moment.

[0062] Therefore, the cost function of the system can be written in the following decomposition form

[0063]

[0064] The above formula satisfies the following constraints

[0065] X i (k)={xi (k | k),x i (k+1 | k),…,x i (k+N-1 | k)}

[0066] U i (k)={u i (k | k),u i (k+1 | k),…,u i (k+N-1 | k)}

[0067] {X j≠i (k)}={X j (k) | j≠i}

[0068] {U j≠i (k)}={U j (k) | j≠i}

[0069] Among them, X i (k) and U i (k) represent the N-step predicted state and N-step predicted control input sequence of the i-th UUV, respectively; {X j≠i (k)} and {U j≠i (k)} then represent the N-step prediction state set of other UUVs and the N-step prediction control input set of other UUVs, respectively; It is the weighting constant; x i (k+N-1 | k) and u i (k+N-1 | k) represent the Nth step prediction state and the Nth step prediction control input sequence of the i-th UUV starting from the current k-time, respectively.

[0070] The overall system optimization problem can be decomposed into N smaller-scale local rolling time-domain optimization problems, each corresponding to a UUV. For the i-th UUV, its local rolling optimization control model is as follows:

[0071]

[0072] The above equation has the following constraints.

[0073] X i (k+q+1|k)=f i (x i (k+q|k),u i (k+q|k)),

[0074] q=0,1,…,N-1;i=1,2,…,N

[0075] x i (k|k)=x i (k)

[0076] ≤0

[0077] where G()≤0 is the constraint condition of the system, is the optimal control input to be solved. It can be seen from the above analysis that, due to the local information interaction between UUVs, the state sequence and the predicted control sequence of the neighborhood UUV can be obtained through local communication.

[0078] Step 2, establish a local communication mode of the group.

[0079] A local communication network of the group is established for local information interaction between UUVs;

[0080] The optimal predicted state sequence and the optimal control sequence of the current UUV obtained by the outer controller are uploaded to the local communication network for use by the outer controller of the neighborhood UUV.

[0081] It should be noted that the multi-UUV cluster hunting control system proposed in the application is composed of multiple independent UUVs, and each hunting UUV is relatively independently controlled and only interacts through a common hunting control task and an upper hunting path constraint.

[0082] Local information interaction is performed between UUVs, and the state sequence and the predicted control sequence of the neighborhood UUV are obtained through local communication. Two-dimensional grid coding processing is applied in the local communication topology to obtain the neighborhood set N i of the i-th UUV, so as to exchange target information and state information.

[0083] Each UUV predicts its own state at the k-th moment to obtain a predicted state sequence, and two-dimensional grid coding processing is applied in the local communication topology to obtain the neighborhood set N i of the i-th UUV, so as to exchange target information and state information.

[0084] The method of introducing state feature coding is used to process the problem of uncertain number of neighborhood nodes. Since each node in the UUV group knows its global coordinates and heading angle and other information, each UUV node is mapped to a two-dimensional grid based on the spatial relationship within the communication range, and the mapping principle is as shown in Figure 3 .

[0085] Figure 3 The communication model based on two-dimensional grid mapping is shown, Figure 3 (a) represents the intuitive mapping of neighborhood relationship in a local group with a UUV as the center and a communication distance as the radius, and a certain angle interval is marked with yellow shadow in the figure to correspond to Figure 3 (b) and Figure 3Mapping relationship in (c), Figure 3 (b) is a deconstruction description of the mapping relationship, Figure 3 (c) is to map the radial communication area in the two-dimensional grid, and the communication area is divided according to the equal interval distance and the direction, and further information interaction is performed by judging whether there is a neighborhood UUV in the interval. The shaded area in the figure can be an encoding area designed for different tasks of the UUV, and the present application takes the heading angle concerned in the pursuit task as the interval division principle, determines the angle interval according to the number of individuals in the group, so as to ensure that the UUV only communicates with at most two individuals in the neighborhood.

[0086] Step 3, the optimal prediction state sequence and the optimal control sequence of the current UUV.

[0087] The prediction state sequence of the current UUV and the optimal prediction state sequence of the neighborhood UUV are obtained by the outer controller, and the optimal prediction state sequence and the optimal control sequence of the current UUV are obtained by optimization solving, wherein the initial optimal prediction state sequence of the neighborhood UUV is obtained based on the initial state of the neighborhood UUV;

[0088] The neighborhood UUV set corresponding to each UUV is obtained by applying two-dimensional grid coding processing in the local communication network to obtain the neighborhood UUV set of the current UUV.

[0089] In this example, the outer DMPC-based pursuit controller obtains the state vector x i (t) of the system state prediction of the current UUV and the optimal prediction state sequence of the neighborhood UUV, and obtains the optimal prediction state sequence and the optimal control sequence by optimization solving, respectively.

[0090] In the optimization solving process, the pursuit control needs to maintain the following three control targets for each UUV:

[0091] (1) taking the moving target as the center, converging the pursuit radius to a fixed expected value;

[0092] (2) the UUVs form a surrounding encirclement with the expected angular velocity around the target;

[0093] (3) the UUVs distributed around the moving target have the same angular interval between each other;

[0094] Wherein, the moving target is the pursuit target.

[0095] The prediction state sequence of the current UUV is obtained according to the following steps:

[0096] Constructing a nonlinear UUV dynamics model equation;

[0097] According to the nonlinear UUV dynamics model equation, the current UUV state is predicted to obtain the predicted state sequence of the current UUV.

[0098] At each time of cyclic solution, each UUV uses the future predicted control input sequence to predict the state of the future M steps. According to the system dynamics model:

[0099]

[0100] The prediction of the system state can be expressed as

[0101]

[0102] wherein, represents the predicted state of the system at the future n th sampling time, represents the predicted control vector at the future n th sampling time, and F is the above-mentioned nonlinear UUV dynamics model equation, is the sampling time.

[0103] The present application designs a specific distributed MPC controller for the pursuit task. The three control targets of the pursuit task are related to the position relationship and angle relationship between the pursuit UUV and the moving target, denoted as is the distance between the pursuit UUV and the moving target, and is specifically calculated as shown in the following formula

[0104]

[0105] The first-order derivative of the envelope angle in the pursuit process and the velocity vector of the pursuit UUV are introduced as the control targets in the pursuit task, wherein the calculation method of the first-order derivative of the envelope angle is as follows

[0106]

[0107] A global cost function is defined, which describes the interaction between all UUVs in the group and the pursuit strategy with the moving target. The angular velocity of the i th UUV around the moving target is denoted as (t), and the angle difference formed by the i th and j th adjacent UUVs relative to the target is denoted as 0,2π] According to the description of the local communication structure in step 2, the UUVs ahead of the i th UUV in the pursuit process are denoted as k, and the UUVs behind the i th UUV are denoted as j. A normal number γ∈[0,1] is introduced as a gain, which can adjust the angle difference between the pursuit UUVs The convergence speed of the desired angular velocity of the ith UUV around the moving target before forming the encircling formation is related to the angular difference between the ith UUV and the UUVs in front and behind it, and is calculated as follows:

[0108]

[0109] Through the above constraint, the angular velocity of each UUV around the target will eventually converge to the desired angular velocity For the multi-UUV cluster hunting control task, the cost function in the distributed MPC algorithm is defined as follows:

[0110]

[0111] When the spacing between two adjacent UUVs is larger than the desired spacing, the speed needs to be controlled to achieve a uniform encircling state. This situation may be contrary to the control instruction given by the second principle of hunting control, so the angular error of the formation during hunting is introduced for adaptive optimization:

[0112]

[0113] where the expected value of the angular difference between adjacent individuals is D = 2π / N, N is the number of hunting UUVs; and ik (t) is the angular difference between the ith UUV and the leading UUV (labeled k) relative to the hunting target, and N1 (t) is the angular difference between the Nth UUV and the leading UUV (the 1st) relative to the hunting target.

[0114] The alternative Lyapunov function is defined as follows:

[0115]

[0116] Let:

[0117]

[0118] When the parameter γ is positive, it can be proved that the above Lyapunov function is positive definite.

[0119] It can be concluded that the desired hunting angular velocity of the ith UUV is

[0120]

[0121] After introducing this adaptive desired angular velocity, the desired hunting angular velocity of the ith UUV and the final desired angular velocity are related as shown in the above formula, and the cost function formula in the model control system finally becomes

[0122]

[0123] where σ denotes the sampling time in the model prediction process, η i,t (t+mσ) denotes the predicted distance between the current UUV and the target at the time of m steps later, then denotes the predicted angular velocity of the current UUV rotating around the target at the time of m steps later. sdand spare constants, which are used for the de-dimensioning of each error in the cost function, and have sd=χ as the expected pursuit distance and sp= as the expected pursuit angular velocity; α, β are both normal numbers, which represent the weights of the control objectives in the cost function. Set N i denotes the set of UUVs in the neighborhood of the current i-th UUV, subscript j denotes the UUVs falling behind the current UUV, and subscript k denotes the UUVs leading the current UUV. Constant α can be regarded as the weight coefficient of the control of the pursuit distance, and constant β can be regarded as the weight coefficient of the convergence speed of the separation angle in the pursuit process.

[0124] Define the prediction step of the controller in each subsystem as M, then the local MPC control strategy can be expressed by the following formula.

[0125]

[0126]

[0127] where M denotes the prediction step of the controller, and σ denotes the sampling time in the model prediction process. The prediction control vector in M steps in the one-time optimization process is denoted as , (iσ;t) denotes the predicted system state at the i-th sampling time in the one-time optimization step, (iσ;t) denotes the system input at the i-th sampling time in the one-time optimization step. The optimization is performed in M prediction steps, and the optimal cost function is obtained, so that the optimal predicted state sequence and the optimal predicted control sequence are obtained, respectively.

[0128] Step 4, the final control quantity of the current UUV.

[0129] The prediction state sequence of the current UUV is subtracted from the actual measured system state quantity through the inner controller to obtain an error, and the error model is defined by the error. The error model is solved to obtain the control compensation quantity of the current UUV;

[0130] The final control quantity of the current UUV is obtained by adding the control compensation quantity of the current UUV to the optimal predicted state sequence and the optimal control sequence of the current UUV;

[0131] In the present embodiment, the current UUV will solve to obtain the optimal predicted state sequence Upload to the local communication network for the DMPC-based controller of the neighborhood UUV to solve the use. At the same time, the optimal predicted state sequence And the optimal control sequence The inner layer Tube-based MPC controller is transmitted to the inner layer Tube-based MPC controller, the inner layer controller, and the state quantity of the actual measurement system is subtracted from the predicted state sequence as the system error input to the error model. The control law design of the error model can be summed up with the control quantity obtained by the distributed controller to obtain the final control output, and then added to the optimal control sequence obtained by the outer layer DMPC controller to obtain the final control output.

[0132] Obtain the control compensation quantity of the current UUV, specifically including:

[0133] A new state vector is defined to describe the motion state of the UUV, and the model is linearized and approximated to obtain a new motion model of the UUV; the predicted state quantity of the nominal system is subtracted from the state quantity of the actual measurement system to obtain an error, which is defined as an error model.

[0134] The feedback gain K of the UUV system in the stable state is selected as the solution Of the linear quadratic regulator of the error model as the control compensation quantity of the current UUV.

[0135] In order to compensate for the inaccuracy of the system and resist the interference of the surface wind wave, the control system is optimized and designed, the Tube idea is introduced, the uncertainty part of the system and the surface wind wave part are regarded as system errors, and the predicted state calculated by the distributed controller is subtracted from the actual measurement state as the system error in the control process.

[0136] First, the feedforward control input in the hunting process is calculated for the system, and the system with feedforward control input and zero external disturbance is called "nominal system" by the present application; then, a state feedback controller is designed to suppress the error between the actual state quantity of the system and the predicted state quantity generated by the nominal system, and the final control quantity of the inner layer control system can be obtained by combining the above two state quantities.

[0137] A new state vector ξ= ∈ is defined to describe the motion state of the UUV, wherein the x and y of the ground coordinate system point to the east and north directions respectively, Ψ represents the heading angle of the UUV, u is the longitudinal velocity, v is the transverse velocity, and r represents the linear velocity of the UUV rotating around the mass point. The nonlinear model is linearized and approximated, and the motion model of a single UUV will be expressed in the following form:

[0138]

[0139] where the system matrix, the nonlinear Lipschitz function and the control matrix are given as follows

[0140]

[0141]

[0142] where m is the mass of the UUV, , , are the added masses, is the moment of inertia in the vertical direction, , , , , , are the first and second order fluid dynamic coefficients respectively. , , , , , .

[0143] For the re-formulated system, the control input of the present application can be divided into two parts. First, the feedforward control input is calculated for the system during the pursuit process, and the system with the feedforward control input and the external disturbance being zero is called the "nominal system"; then a state feedback controller is designed to suppress the error between the actual state quantity of the system and the predicted state quantity of the nominal system, and in addition, the two state quantities are combined to obtain the final control quantity of the inner control system.

[0144] The N-step control sequence and the N-step disturbance sequence of the system are respectively denoted as τ={ , ,…, } and ω={ , ,…, }, and the discrete-time linearized nominal model of the system can be expressed in the following form:

[0145]

[0146] The difference between the predicted state quantity of the nominal system and the actual measured state quantity of the system is denoted as , and the control input is divided into the following two parts according to the foregoing description wherein, : R n → R m , and the error model is defined in the following expression:

[0147]

[0148] The feedback gain K of the stable state of the UUV system is selected as the solution of the linear quadratic regulator of the error system in the above formula The controller K is used for calculating the robust invariant set, and the robust invariant set is calculated offline, so that the overall control output becomes:

[0149]

[0150] Step 5, update of the optimal control.

[0151] After a control period, the current UUV obtains the final control quantity of the next period through the outer controller and the inner controller again, and the current UUV is controlled by hunting.

[0152] After a control period, the current UUV obtains the final control quantity of the next period through the outer controller and the inner controller again, and the current UUV is controlled by hunting.

[0153] After a control period, the current UUV obtains the final control quantity of the next period through the outer controller and the inner controller again, and the current UUV is controlled by hunting.

[0154] In the present example, after the optimal control sequence is obtained by the presequence calculation of the actuator control, the control vector containing N steps is obtained, the first item in the sequence is used as the solution, and at the next sampling time, step 3 is entered to use the updated state quantity as the initial condition and perform optimization again.

[0155] It should be noted that the model predictive control adopts a rolling finite time domain optimization strategy, and the UUV solves the optimal control rate in the finite time domain starting from the current time according to the optimization performance index at the current time, obtains the optimal control input sequence, and uses the first item in the sequence as the solution to the system. At the next sampling time, use the updated state quantity as the initial condition and perform optimization again. In this way, the actual optimal control input sequence can be obtained by continuously solving in a loop.

[0156] In order to further illustrate the UUV hunting method based on distributed model predictive control provided by the present application, simulation verification shows that the double-layer control framework proposed can better reduce the control error and improve the robustness of the system, and the designed distributed hunting control algorithm can effectively and stably form a surrounding situation for the irregularly moving target.

[0157] The irregular motion of the moving target is verified in the simulation verification, and the moving target is assumed to be at the initial position =[-39, 42], the initial position of the UUV group is also randomly generated, the control period σ is set to 0.1 s, the prediction step length M of the controller is 50, and the simulation results are shown in Figs. Figure 4 、 5 、6、7.

[0158] It can be seen from Figure 4 、 Figure 5 that when the moving target makes irregular variable speed motion, in the case of known target position, the distributed MPC encirclement control algorithm proposed in the application can also achieve uniform encirclement of the moving target. It can be seen from Figure 5 that the multi-UUV group also converges the encirclement radius error within ±0.2 m within 40 s, and the uniform distribution formation formed by cooperative encirclement is maintained until the end of the task time, and the distance error is periodically changed.

[0159] Figure 6 and Figure 7 respectively show two control targets in the encirclement task, Figure 6 show the angle interval of each UUV and its subsequent neighbor in the encirclement process, and after forming a uniformly distributed configuration on the encirclement circle, the angle interval between adjacent UUVs is = 360 / N = 72°, and the angle interval converges to 72°±2°. Figure 7 then show the angular velocity variation of each UUV in the encirclement task, which makes an envelope motion around the moving target in a clockwise direction, the simulation sets the expected angular velocity to be 0.09 rad / s, and Figure 7 it can be seen that the angular velocity error periodically changes within ±0.03 rad / s.

[0160] The application provides a UUV encirclement system based on a distributed model predictive control, comprising:

[0161] A control system construction module is used to establish a distributed control system structure, including an inner controller and an outer controller corresponding to each UUV;

[0162] A control sequence calculation module is used to obtain the predicted state sequence of the current UUV and the optimal predicted state sequence of the neighborhood UUV through the outer controller, and to obtain the optimal predicted state sequence and the optimal control sequence of the current UUV through optimization solution;

[0163] A control compensation module is used to obtain the error by subtracting the predicted state sequence of the current UUV from the state quantity of the actual measurement system through the inner controller, and to obtain the control compensation quantity of the current UUV by solving the error model defined by the error.

[0164] The control module is used for adding the current UUV control compensation quantity to the optimal predicted state sequence and the optimal control sequence of the current UUV to obtain a current UUV final control quantity; the current UUV obtains the final control quantity of the next period through the outer controller and the inner controller again after one control period, and is controlled by the pursuit control.

[0165] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for UUV containment based on distributed model predictive control, characterized in that, Used to control multiple UUVs to surround and capture a moving target, including: Establish a distributed control system architecture, including an inner controller and an outer controller for each UUV; The predicted state sequence of the current UUV and the optimal predicted state sequence of the neighboring UUV are obtained by the outer controller, and the optimal predicted state sequence and the optimal control sequence of the current UUV are obtained by optimization. The error is obtained by subtracting the predicted state sequence of the current UUV from the state variables of the actual measurement system through the inner controller. Then, the error model is defined by the error model and the control compensation amount of the current UUV is obtained by solving the error model. The final control quantity of the current UUV is obtained by adding the current UUV control compensation quantity to the current UUV's optimal predicted state sequence and optimal control sequence; After one control cycle, the UUV obtains the final control quantity for the next cycle through the outer and inner controllers to control the UUV for capture.

2. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, The outer controller for each UUV is responsible for target detection, inter-neighborhood information exchange, and solving for the optimal prediction state and optimal control sequence. The inner controller for each UUV uses the optimal predicted state output by the outer controller as a reference state to perform error compensation on the current UUV and obtain the control compensation amount.

3. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, Also includes: Establish a group local communication network for local information exchange among UUVs; The optimal predicted state sequence and optimal control sequence of the current UUV are obtained by the outer controller and uploaded to the local communication network for use by the outer controller of the neighboring UUV.

4. The UUV containment method based on distributed model predictive control according to claim 3, characterized in that, The neighborhood UUV set corresponding to each UUV is obtained by applying two-dimensional grid coding processing in the local communication network to obtain the neighborhood UUV set of the current UUV.

5. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, The predicted state sequence of the current UUV is obtained according to the following steps: Constructing nonlinear UUV dynamic model equations; The current state of the UUV is predicted based on the nonlinear UUV dynamic model equations to obtain the predicted state sequence of the current UUV.

6. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, During the optimization process, the following three control objectives will be maintained for each UUV: (1) Using the moving target as the center, converge the encirclement radius to a fixed expected value; (2) The UUV surrounds the target with the desired angular velocity; (3) The UUVs distributed around the moving target have the same angular intervals in pairs; Among them, the moving target is the target of encirclement and capture.

7. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, After one control cycle, the UUV obtains the final control input for the next cycle through both the outer and inner controllers, including: Currently, after one control cycle, the UUV contains control vectors of N steps. In each control cycle, only the optimal control vector of the first step is selected. After the controller outputs to the actuator, after one control cycle, the final control quantity for the next cycle is obtained again through the outer controller and the inner controller.

8. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, Solving the error model yields the control compensation amount for the current UUV, including: The feedback gain K of the UUV system in steady state is selected as the solution of the linear quadratic regulator of the error model, and is used as the control compensation quantity of the current UUV.

9. The UUV containment method based on distributed model predictive control according to claim 1, characterized in that, The outer controller is a DMPC-based containment controller; the inner controller is a Tube-based MPC controller.

10. A UUV containment system based on distributed model predictive control, characterized in that, include: The control system construction module is used to establish a distributed control system structure, including the inner controller and outer controller corresponding to each UUV; The control sequence calculation module is used to obtain the predicted state sequence of the current UUV and the optimal predicted state sequence of the neighboring UUV through the outer controller, and to perform optimization to obtain the optimal predicted state sequence and the optimal control sequence of the current UUV respectively. The control compensation module is used to obtain the error by subtracting the predicted state sequence of the current UUV from the state quantity of the actual measurement system through the inner controller, and then to define an error model through the error and solve the error model to obtain the control compensation amount of the current UUV. The control module is used to add the current UUV control compensation quantity to the current UUV's optimal predicted state sequence and optimal control sequence to obtain the current UUV's final control quantity; After one control cycle, the UUV obtains the final control quantity for the next cycle through the outer and inner controllers to control the UUV for capture.

Images