A heterogeneous unmanned system data-driven cooperative output regulation method and device

Abstract

The present application belongs to the field of cluster unmanned system cooperative control technology, and particularly relates to a heterogeneous unmanned system data-driven cooperative output regulation method and device. The specific process of the method is as follows: S100, the heterogeneous unmanned system is composed of multiple unmanned units, and follows an external system; an inner membrane system is introduced, and an extended system of each unmanned unit is established according to the dynamic model of the unmanned unit; S200, open-loop trajectory data of the extended system of each unmanned unit is obtained through offline experiments, and a data matrix is obtained; S300, a distributed data-driven control gain matrix is calculated based on the data matrix; S400, a distributed dynamic state feedback controller is obtained based on the state of the inner membrane system and the control gain matrix, and output regulation of the unknown model heterogeneous unmanned system is realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of cooperative control of cluster unmanned systems, and particularly relates to a data-driven cooperative output regulation method and device for heterogeneous unmanned systems. BACKGROUND

[0002] With the interwoven advancement of the new round of scientific and technological revolution, industrial revolution and new military revolution, unmanned systems represented by unmanned vehicles, unmanned aerial vehicles, unmanned ships, etc. have become the best entry point and important starting point for the research of artificial intelligence theory, methods and technology. Heterogeneous cluster unmanned systems are composed of a large number of unmanned system units that interact with each other and have different dynamic characteristics. Through mutual communication, cooperation, competition and other ways, each unmanned system in the heterogeneous cluster unmanned system can complete complex tasks that cannot be completed by a single unmanned system, and realize the leap-forward improvement of productivity and efficiency. In view of the application requirements of tracking a trajectory signal or suppressing a specific type of interference for heterogeneous cluster unmanned systems, it has certain theoretical and practical value to study the cooperative output regulation problem of heterogeneous cluster unmanned systems. In addition, the systems in the heterogeneous cluster unmanned system are strongly coupled, the cooperative mechanism is complex, and the external complex environment leads to incomplete and uncertain system information, which brings great challenges to the research of the cooperative output regulation problem of cluster systems and attracts widespread attention from scholars.

[0003] However, the existing research on the cooperative output regulation problem of heterogeneous cluster unmanned systems all adopts the model-based control method, that is, the mathematical model of the controlled object needs to be established based on the system mechanism or identification method. However, with the increase of the scale of the heterogeneous unmanned cluster system and the complexity of the unit model, obtaining an accurate mathematical model becomes an important bottleneck that limits the design of distributed control strategy and the realization of control objectives in complex environments. In recent years, data-driven control methods have emerged, that is, only data is used to realize the control and analysis of unknown systems. For the data-driven control problem of a single unknown system, scholars have carried out in-depth and detailed research. However, for the cooperative output regulation problem of multi-body systems, due to the coupling noise in the data collected in the actual scene, the method of solving the output regulation equation to achieve the control objective can only achieve approximate output regulation, that is, it cannot achieve zero-deviation tracking. This is because the noise in the data will cause countless system models to match the collected data, resulting in no solution to the output regulation equation that satisfies countless systems, and it is impossible to construct a data-driven output regulation controller.

[0004] Therefore, how to use noisy data to achieve the cooperative output regulation of heterogeneous cluster unmanned systems with unknown models in a distributed structure to complete the planning and design of control strategies and ultimately achieve the control objective of zero-deviation cooperative output regulation has become a technical problem to be solved. SUMMARY

[0005] To solve the above problems, the application considers the differences and noise data caused by the structure and dynamic performance of different types of unmanned single bodies, and proposes a data-driven cooperative output regulation method and device for heterogeneous unmanned systems, which can realize output zero deviation tracking of unknown model heterogeneous unmanned systems and external systems.

[0006] In a first aspect, the application provides a data-driven cooperative output regulation method for heterogeneous unmanned systems, and the specific process is as follows:

[0007] S100, the heterogeneous unmanned system is composed of multiple unmanned single bodies, and the heterogeneous unmanned system follows an external system; an inner membrane system is introduced, and an extended system of each unmanned single body is established according to the dynamic model of the unmanned single body;

[0008] S200, the open-loop trajectory data of the extended system of each unmanned single body is obtained through offline experiments, and a data matrix is obtained;

[0009] S300, based on the data matrix, a distributed data-driven control gain matrix is calculated;

[0010] S400, based on the state of the inner membrane system and the control gain matrix, a distributed dynamic state feedback controller is obtained, and the output regulation of the unknown model heterogeneous unmanned system is realized.

[0011] Further, the inner membrane system of the application is:

[0012] The dynamic model of the external system is:

[0013]

[0014] y0(t)=-Fv(t)

[0015] wherein v(t)、 and y0(t) are the state, state derivative and output of the external system, S and F are known real matrices; the above heterogeneous unmanned system and the external system can be regarded as an unmanned system with a leader-following structure, wherein the external system is the leader, and the N unmanned single bodies are the followers, the communication network between them is described by graph , which contains a directed spanning tree with the leader as the root node.

[0016] The inner membrane system is:

[0017]

[0018] wherein z i (t) and are the state and state derivative of the inner model system, e vi (t) is a virtual tracking error; ​

[0019]

[0020] where β is an n β × n β constant matrix whose characteristic polynomial is the same as the minimal polynomial of matrix S, σ is an arbitrary chosen n β × 1 constant column vector, (β×σ) is controllable, blockdiag(...) represents a block diagonal matrix constructed with the sub-matrices on the diagonal; n y -tuple represents that the block diagonal matrix has n y sub-matrices.

[0021] Further, the virtual tracking error of the application is:

[0022]

[0023] where a ij is the element of the adjacency matrix of graph , a i0 represents the connection relationship between unmanned body i and external system, y i (t), y j (t) and y0(t) are respectively the outputs of unmanned body i, unmanned body j and external system.

[0024] Further, the application establishes an extended system of each unmanned body according to the dynamic model of the unmanned body, and the specific process is:

[0025] Suppose the dynamic model of each unmanned body is:

[0026]

[0027] y i (t)=C i x i (t), i=1, 2,..., N

[0028] where x i (t) and are the state and state derivative of unmanned body i, u i (t) is the control input of unmanned body i, y i (t) is the output of unmanned body i, v(t) is the state of external system, and matrices A i , B i , E i and C i are unknown real matrices.

[0029] The extended system of each unmanned body is established as:

[0030]

[0031] in,

[0032] Furthermore, the present invention introduces noise d in the expansion system i (t), run the extended system in open loop and obtain the data matrix.

[0033] Furthermore, the data matrix obtained in the present invention is specifically: applying the control input U at T time points i- , the control input satisfies n xi +n z +1-order continuous excitation, n ui is the dimension of the control input of the unmanned unit i, n z is the state dimension of the internal model system, collecting the state X of the unmanned unit i i- and the state Z of the endomembrane system i- , the calculated state derivative X of the unmanned cell i i+ and the derivative Z of the endomembrane system state i+ ; Construct the matrix as follows:

[0034] U i- =[u i (t1)u i (t2)…u i (t T )]

[0035] X i- =[x i (t1)x i (t2)…x i (t T )]

[0036] Z i- =[z i (t1)z i (t2)…z i (t T )]

[0037]

[0038] Furthermore, the T of the present invention is xi +n z +1)×n ui +n xi +n z -1, where n xi is the state dimension of the unmanned unit i, n ui is the dimension of the control input of the unmanned unit i, n zState dimension of the inner model system.

[0039] Further, for step S300, the distributed data-driven control gain matrix is calculated based on the data matrix, specifically:

[0040] Solve the following matrix inequality for the unmanned monomer i:

[0041]

[0042] Wherein, Y i and P i are variables to be solved. is the maximum upper bound of the noise introduced by the unmanned monomer i in the data collection stage, and I represents a unit matrix with appropriate dimension;

[0043] The control gain matrix is designed as K i =Y i P i -1 / λ i , wherein λ i is the ith eigenvalue of the matrix H, and the matrix is the Laplace matrix of the connected graph composed of the external system and the unmanned system.

[0044] Further, in step S400 of the present application, the distributed dynamic state feedback controller is designed as:

[0045]

[0046] Wherein, a ij is the element in the ith row and jth column of the adjacency matrix , and a i0 represents the connection relationship between the unmanned monomer i and the external system.

[0047] In a second aspect, the application provides a data-driven cooperative output regulation device for a heterogeneous unmanned system, comprising the steps of:

[0048] An extended system construction module is provided, wherein the heterogeneous unmanned system is composed of multiple unmanned monomers, which follow an external system; an inner membrane system is introduced, and an extended system of each unmanned monomer is established according to the dynamic model of the unmanned monomer;

[0049] A data acquisition module is provided, which obtains the open-loop trajectory data of the extended system of each unmanned monomer through offline experiments, and obtains a data matrix;

[0050] a gain matrix calculation module, which calculates a distributed data-driven control gain matrix based on the data matrix;

[0051] a control output module, which obtains a distributed dynamic state feedback controller based on the inner membrane system state and the control gain matrix, so as to realize output adjustment of the unknown model heterogeneous unmanned system.

[0052] Advantages

[0053] First, the present application only uses open-loop collected state and input data to design a distributed controller, eliminating the dependence of traditional control on system models.

[0054] Second, the distributed data-driven control gain matrix design method proposed by the present application only needs to solve a low-complexity linear matrix inequality offline, and does not need to solve optimization problems online, thereby saving computing resources and energy consumption to a certain extent.

[0055] Third, the data-driven collaborative output adjustment method for a heterogeneous unmanned system proposed by the present application effectively solves the collaborative output adjustment problem under the condition of data interference, and for the first time realizes zero-deviation tracking of the output of each unmanned single unit in the system and the external system output, and has good robustness. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0057] Figure 1 is a flowchart of a data-driven collaborative output adjustment method and device for a heterogeneous unmanned system provided by the present application;

[0058] Figure 2 is a collaborative output adjustment effect diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The embodiments of the present application will be described in detail below with reference to the drawings.

[0060] It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict; and all other embodiments obtained by those skilled in the art on the basis of the embodiments in the present disclosure without creative labor are within the scope of protection of the present disclosure.

[0061] It is important to note that the various aspects described throughout this disclosure can be combined in a wide variety of ways. It should be apparent that aspects described herein can be implemented in various forms of hardware, software, or a combination thereof; and that the specific design and implementation details are not limiting, but are merely exemplary. Based on the disclosure provided herein, one of ordinary skill in the art will be able to devise numerous ways to implement the aspects described herein while staying within the scope of the appended claims.

[0062] As Figure 1 shown in the flowchart of the present application, a heterogeneous unmanned system data-driven cooperative output regulation method comprises the following steps:

[0063] S100, a heterogeneous unmanned system is composed of multiple unmanned units, which follow an external system; an internal membrane system is introduced, and an extended system of each unmanned unit is established according to the dynamic model of the unmanned unit.

[0064] In one embodiment, the heterogeneous unmanned system described in step S100 is composed of N unmanned units with different dynamic models, and the dynamic model of each unmanned unit is as follows:

[0065]

[0066] y i (t) = C i x i (t), i = 1, 2,..., N

[0067] wherein, at time t > 0, the state x i (t) of the unmanned unit i and the derivative dimension of the state are n xi , the dimension of the control input u i (t) of the unmanned unit i is n ui , the dimension of the output y i (t) of the unmanned unit i is n y , and the dimension of the state v(t) of the external system is n v , which generally includes a reference signal to be tracked and an interference signal to be suppressed. The matrix A i is an unknown real matrix of n xi × n xi dimension, the matrix B i is an unknown real matrix of n xi × n ui dimension, and the matrix E i is an unknown real matrix of n xi × n vunknown real matrix of dimension n i is an n y ×n xi unknown real matrix of dimension n

[0068] The dynamics of the external system is given by:

[0069]

[0070] y0(t) = -Fv(t)

[0071] where the derivative of the state of the external system and the output y0(t) are of dimension n v and n y respectively, S is an n v ×n v known real matrix with no eigenvalue with negative real part, and F is an n y ×n v known real matrix.

[0072] The heterogeneous unmanned system and the external system can be unified as an unmanned system with a leader-follower structure, where the external system is the leader and the N unmanned agents are the followers. The communication network between them can be described by a graph which contains a directed spanning tree with the leader as the root node.

[0073] Based on the definition of the dynamics of the external system and the unmanned agent system, the extended closed-loop system in step S100 is established by the following steps:

[0074] S111, introduce the internal membrane system:

[0075]

[0076] where the state z i (t) and the derivative of the state of the internal membrane system are of dimension n z . Matrices G1 and G2 are internal membrane matrices related to the matrix S of the external system, and are designed as:

[0077]

[0078] where β is a constant matrix of dimension n β ×n β with the same minimal polynomial as the matrix S, blockdiag(β,...,β) represents a block diagonal matrix with β as the diagonal submatrix, and n y -tuple indicates that the block diagonal matrix has n y ​sub-matrix, σ is an arbitrary selected constant column vector of dimension n β ×1, as long as (β×σ) is controllable.

[0079] e vi (t) is the virtual tracking error, designed as:

[0080]

[0081] where a ij is the adjacency matrix of the graph , the element in the i-th row and j-th column of the adjacency matrix , a i0 represents the connectivity relationship between the unmanned body i and the external system, y i (t), y j (t) and y0(t) are the outputs of the unmanned body i, unmanned body j and the external system, respectively.

[0082] S112, define the extended vector with dimension n ξi =n xi +n z , thus the extended system is constructed as:

[0083]

[0084] where,

[0085] S200, collect the open-loop trajectory data of the extended system of each unmanned body through offline experiments, obtain the input data matrix, the body state matrix and the inner membrane system input matrix.

[0086] In one embodiment, the data matrix in step S200, the construction step is:

[0087] S211, introduce noise d i (t) to the extended system in step S100, run the above extended system in open loop, collect noise data:

[0088]

[0089] Specifically, for any time period 0 to T1, randomly generate the input to the unknown unmanned body i in real time, obtain the state of the unmanned body and the inner membrane state in this period. In any sampling T time points in the time period 0 to T1, satisfy 0 T ≤T1, T=(n xi +n z +1)×n ui +n xi +n z-1. If the input sequence corresponding to the T time points is a sustained excitation of order n xi + n z + 1, then record these time points and this sequence input. Record the state data of the unmanned body i at the corresponding time Endomembrane state data When recording the state of the unmanned body i and the state of the endomembrane system at time τ, record the state data x i (τ-Δt), x i (τ+Δt), z i (τ-Δt) and z i (τ+Δt) of the two time points close to it before and after, where Δt is as close to zero as possible. Estimate the derivative of the state of the unmanned body i at each time τ as The derivative of the state of the endomembrane system is

[0090] S212, using the applied control input, the collected state of the unmanned body i, the collected state of the endomembrane system, the calculated derivative of the state of the unmanned body i, and the calculated derivative of the state of the endomembrane system, construct the input data matrix U i- of the unmanned body i, the state data matrix X i- of the unknown physical process, the state data matrix Z i- of the endomembrane system, the derivative data matrix X i+ of the state of the unknown physical process, and the derivative data matrix Z i+ of the state of the endomembrane system as follows:

[0091] U i- = [u i (t1) u i (t2) … u i (t T )]

[0092] X i- = [x i (t1) x i (t2) … x i (t T )]

[0093] Z i- = [z i (t1) z i (t2) … z i (t T )]

[0094]

[0095] S300, based on the data matrix, calculate a distributed data-driven control gain matrix.

[0096] In one embodiment, the distributed data-driven control gain matrix design method in step S300, the specific design steps are:

[0097] S311, set a given normal number As the maximum upper bound of the noise introduced by the unmanned single body i in the data collection stage, this noise contains the derivative estimation error and the external signal in the data collection stage, the following matrix is constructed using the data matrix constructed in step S200: augmented state data matrix Augmented state derivative data matrix And

[0098]

[0099] Where I represents a unit matrix with appropriate dimensions.

[0100] S312, based on the above data matrix, solve the following matrix inequality for each unmanned single body i:

[0101]

[0102] Where Y i and P i are to be solved variables, matrix P i is a positive definite real symmetric matrix of n ξi ×n ξi dimension, matrix Y i is a matrix of n ui ×n ξi dimension. Thus, the control gain matrix is designed as K i =Y i P i -1 / λ i , where λ i is the ith eigenvalue of matrix H, matrix is the Laplacian matrix of graph , Λ is a diagonal matrix composed of a i0 , i = 1, 2, …, N, that is Where a i0 represents the connection relationship between the unmanned single body i and the external system, if the unmanned single body i can obtain the information of the external system, then a i0 = 1, otherwise 0.

[0103] S400, based on the membrane system state in S100 and the control gain matrix obtained in S300, obtain a distributed dynamic state feedback controller to realize the output regulation of the unknown model heterogeneous unmanned system.

[0104] In one embodiment, the distributed dynamic state feedback controller in step S400 is designed as:

[0105]

[0106] In the online running phase, each unknown model unmanned body i sends the current state x i (t) to the above-mentioned distributed dynamic state feedback controller at each moment, and the inner membrane system sends the current inner membrane state z i (t) to the distributed dynamic state feedback controller, the above-mentioned distributed dynamic state feedback controller generates the control input u i (t) and sends it back to the unknown unmanned body i, so as to realize that the output y i (t) of the unknown unmanned body i asymptotically tracks the output y0(t) of the external system, that is, the output regulation of the unknown model heterogeneous unmanned system is realized.

[0107] In one embodiment, the feasibility and effectiveness of the data-driven output synchronization strategy design method in the application are verified by simulation experiments.

[0108] As Figure 2 shown is an effect diagram of the method in the application running on a multi-robot system for 50 seconds in one embodiment. The corresponding system matrix is:

[0109]

[0110] Wherein, i = 1, 2, 3, 4, ω1 = π / 5, ω2 = 1. According to step S100, the inner membrane matrices G1 and G2 are designed as

[0111]

[0112] The selected parameter T = 20, The four subgraphs in the figure respectively represent the cooperative output regulation effect diagrams of the four robots, and the solid line in each subgraph is the expected tracked external system output, the dashed line is the output of the robot, and the dotted line is the tracking error. It can be seen that the output of each robot can track the output of the external system without deviation, that is, the output regulation of the multi-robot system is realized, and the effectiveness of the proposed data-driven cooperative output regulation method is proved.

[0113] The above is only a specific embodiment of the application, but the protection scope of the application is not limited thereto, and any changes or replacements within the technical range disclosed in the application can be easily thought of by those skilled in the art, which should be covered in the protection scope of the application. Therefore, the protection scope of the application should be subject to the protection scope of the claims.

Claims

1. A heterogeneous unmanned system data-driven collaborative output regulation method, characterized in that, The method comprises the steps of: S100, assuming that a heterogeneous unmanned system is composed of a plurality of unmanned units, the heterogeneous unmanned system follows an external system; introducing an inner membrane system, and establishing an extended system of each unmanned unit according to a dynamic model of the unmanned unit; S200, obtaining open-loop trajectory data of the extended system of each unmanned unit through offline experiments, and obtaining a data matrix; S300, calculating a distributed data-driven control gain matrix based on the data matrix; S400, obtaining a distributed dynamic state feedback controller based on a state of the inner membrane system and the control gain matrix, and realizing output adjustment of the heterogeneous unmanned system; The inner membrane system is: The dynamic model of the external system is: y0(t) = -Fv(t) Wherein, v(t) and y0(t) is the state, state derivative, and output of the external system, S and F are known real matrices; The inner membrane system is: where z i (t) and are the state and state derivative of the inner model system, e vi (t) is the virtual tracking error; where β is an n β x n β constant matrix whose characteristic polynomial is the same as the minimal polynomial of the matrix S, σ is an arbitrary chosen constant column vector of dimension n β x 1, and blockdiag(...) denotes a block-diagonal matrix constructed with the sub-matrices on the diagonal; n y -tuple indicates that the block-diagonal matrix has n y sub-matrices; The virtual tracking error is: wherein a ij is a graph of adjacency matrix of the ith row and jth column, a i0 represents the communication relationship between the unmanned cell i and the external system, y i (t), y j (t) and y0(t) are the outputs of the unmanned cell i, the unmanned cell j and the external system respectively; the graph is a communication network graph describing the heterogeneous unmanned system and the external system; According to the dynamic model of the unmanned unit, the extended system of each unmanned unit is established, and the specific process is: The dynamic model of each unmanned unit is: y i (t) = C i x i (t), i = 1, 2,..., N where x i (t) is the state of the unmanned monomer i, and is the state derivative of the unmanned monomer i, u i (t) is the control input of the unmanned monomer i, y i (t) is the output of the unmanned monomer i, v(t) is the state of the external system, and matrices A i , B i , E i , and C i are real matrices; The extended system of each unmanned unit is established as: wherein, 2. The method of claim 1, wherein, The extended system introduces noise d i (t), in the open loop running the extended system, obtaining a data matrix.

3. The data-driven collaborative output regulation method for heterogeneous unmanned systems according to claim 2, wherein, The obtained data matrix is specifically: applying control input U at T time points i- , the control input satisfies n xi +n z +1 order sustained excitation, n ui is the dimension of the control input of the unmanned single body i, n z is the state dimension of the internal model system, the state X i- of the unmanned single body i and the state Z i- of the internal model system are collected, the calculated state derivative X i+ of the unmanned single body i and the state derivative Z i+ of the internal model system are calculated; the matrix is constructed as follows: U i- = [u i (t1)u i (t2)…u i (t T )] X i- = [x i (t1)x i (t2)…x i (t T )] Z i- = [z i (t1)z i (t2)…z i (t T )] 4. The data-driven collaborative output regulation method for heterogeneous unmanned systems according to claim 3, wherein, The T = (n xi + n z + 1) x n ui + n xi + n z - 1, where n xi is the state dimension of the unmanned monobloc i, n ui is the dimension of the control input of the unmanned monobloc i, and n z is the state dimension of the internal model system.

5. The data-driven collaborative output regulation method for heterogeneous unmanned systems according to claim 3, wherein, For step S300, the distributed data-driven control gain matrix is calculated based on the data matrix, and specifically: Solving the following matrix inequality for the unmanned unit i: wherein Y i and P i are variables to be solved; is the maximum upper bound of the noise introduced by the individual i in the data collection phase, I denotes a dimension-appropriate identity matrix; The control gain matrix is ​​designed as K i =Y i P i -1 / λ i , where λ i is the i-th eigenvalue of matrix H, the matrix It is a connected graph consisting of external systems and unmanned systems The Laplace matrix of .

6. The data-driven collaborative output regulation method for heterogeneous unmanned systems according to claim 1, wherein, Step S400, the distributed dynamic state feedback controller is designed as: where K i is a control gain matrix, a ij is an adjacency matrix whose element in the i-th row and j-th column is a i0 denotes the connectivity relationship of the unmanned monomer i and the external system.

7. A heterogeneous unmanned system data-driven collaborative output regulation apparatus for implementing the heterogeneous unmanned system data-driven collaborative output regulation method of claim 1, wherein Comprise: The extended system construction module assumes that a heterogeneous unmanned system is composed of a plurality of unmanned units, the heterogeneous unmanned system follows an external system; introducing an inner membrane system, and establishing an extended system of each unmanned unit according to a dynamic model of the unmanned unit; The data acquisition module obtains open-loop trajectory data of the extended system of each unmanned unit through offline experiments, and obtains a data matrix; The gain matrix calculation module calculates a distributed data-driven control gain matrix based on the data matrix; The control output module obtains a distributed dynamic state feedback controller based on a state of the inner membrane system and the control gain matrix, and realizes output adjustment of the heterogeneous unmanned system.

Images