A distributed filter-based iterative learning target trajectory tracking method

By using distributed filters in iterative learning control, the problem of slow trajectory tracking response speed and insufficient robustness in non-regular systems is solved, and information interaction and trajectory convergence between multiple controlled systems is achieved, which improves the System robustness and tracking efficiency.

CN113934140BActive Publication Date: 2024-11-22WUXI ELECTRONICS & INSTR IND
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010677619.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-07-14
Publication Date
2024-11-22
Estimated Expiration
2040-07-14

AI Technical Summary

Technical Problem

The existing iterative learning control methods are difficult to achieve fast tracking response and robust trajectory tracking in non-regular systems, especially in the information interaction and collaborative control between multiple controlled systems.

Method used

Using the iterative learning target trajectory tracking method based on distributed filters, the state estimation and control volume interaction between multiple controlled systems is realized through offline calculation and information fusion of distributed filter gain, and the sparse matrix and positive definite matrix R are used. Perform system parameter optimization, design full-order filters and distributed iterative learning law to achieve consistent trajectory convergence.

Benefits of technology

Improve the information interaction and tracking response speed between controlled systems, enhance the robustness of the system, realize the tracking capability of multiple controlled systems, and has faster response speed and Stronger stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113934140B_ABST
    Figure CN113934140B_ABST
Patent Text Reader

Abstract

The present invention relates to an iterative learning target trajectory tracking method based on distributed filters, which belongs to the field of communication information processing. It constructs an iterative learning target trajectory tracking method based on distributed filters for non-regular systems. The iterative learning target trajectory tracking method based on distributed filters designed by the present invention is based on distributed filters and can realize trajectory tracking of multiple controlled systems at the same time. Compared with a single filter iterative learning control system, an iterative learning control system using a distributed filter has the characteristics of information interaction of estimated quantities and control quantities between controlled systems, fast tracking response speed, strong robustness, etc.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to an iterative learning target trajectory tracking method based on a distributed filter, and belongs to the field of communication information processing. Background Art

[0002] Iterative learning control (ILC) is suitable for situations where operations are repeated in a limited interval. It uses previous control inputs and outputs to generate the current input in order to improve the output effect. After multiple iterations, the system can achieve completely error-free tracking. The goal of iterative learning control is to achieve complete tracking in a limited interval. Due to its simple design, small amount of online calculations, and good control effect, iterative learning control is used in industrial robot control, chemical process control, and other occasions. Summary of the invention

[0003] The present invention is based on a distributed filter and constructs an iterative learning target trajectory tracking method based on a distributed filter for a non-regular system. The steps are as follows:

[0004] Step 1: Given the system parameters of the controlled object, the distributed filter gain is calculated by and Perform offline calculations;

[0005] Considering n filter nodes, the estimated error vector at time k is defined System state vector Output estimated error vector and the system parameter vector

[0006]

[0007] definition if but and is a sparse matrix and can be expressed as in

[0008]

[0009] Given γ>0, if there exists a positive definite matrix R, Q=diag{Q1,Q2,…,Q n}, and matrices X, Y, and

[0010] The gain of the filter is

[0011] Step 2: Given the desired tracking trajectory signal of each controlled system, calculate Consider the following discrete-time linear time-invariant controlled system i: in: is an unknown system state, is the system interference, belonging to l2[0,∞), t∈[0,∞) is the iteration time, k is the number of iterations,

[0012] For all t∈[0,T], when k→∞, Converge to in is the system output value of the kth iteration of the controlled object, m is the number of neighboring nodes of node i, and the coefficient matrices A, B, M, and C i , D i All have corresponding dimensions:

[0013] Step 3: Calculate the deviation between the output signal of the controlled system and the signal expected to be tracked

[0014] Step 4: Select the Satisfaction Suitable parameters for And set the number of iterations k;

[0015] Step 5: Calculate the state estimate of the controlled object at the kth iteration. This estimate is based on the distributed filter and simultaneously fuses the state estimates of its neighboring nodes.

[0016] calculate

[0017] Step 6: Information Fusion

[0018] Step 7: Calculate the output of the controlled object at the kth iteration

[0019] The iterative learning target trajectory tracking method based on distributed filters designed by the present invention is based on distributed filters and can realize trajectory tracking of multiple controlled systems at the same time. Compared with the single filter iterative learning control system, the distributed filter iterative learning control system has the characteristics of information interaction of estimated quantities and control quantities between controlled systems, fast tracking response speed, strong robustness, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The network structure diagram of the distributed filter iterative learning control system DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and beneficial effects of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings.

[0022] Assume that there are n controlled objects in total, and the trajectory that each controlled object is expected to track is. At the same time, each controlled object establishes a filter. In the network composed of these controlled objects, each filter will communicate with other controlled objects through a fixed network topology, share local state estimates, and send the control quantity calculated by the local iterative learning law to the neighboring control system. Each controlled object tracks the expected trajectory, not only based on its own information, but also integrating the information provided by the neighboring control systems.

[0023] Figure 1 The structure diagram of the iterative learning multi-control system based on distributed filters. The iterative learning control algorithm shown in the figure uses the filtered output error signal instead of the filtered output signal. k (t) is the disturbance signal that causes the output to deviate from the given value. At the same time, the control quantity calculated by each controlled system through the distributed filter will be transmitted to other controlled systems through the network topology.

[0024] Consider a control system network with n nodes, the topology of which is a directed graph To indicate that is a node set, is the boundary set, is the adjacency matrix. If a directed graph There is an edge from node j to node i, then the ordered pair And a ij > 0, and node j is called an adjacent node of node i. In addition, we assume that for all a ii =1. Therefore, (i, i) can be considered as an additional boundary. The set of all the adjacent nodes of node i plus itself is called the adjacent node set of node i, denoted as

[0025] Consider the following discrete-time linear time-invariant controlled system i:

[0026]

[0027] in: is an unknown system state, is the system disturbance, belonging to l2[0,∞), t∈[0,∞) is the iteration time, and k is the number of iterations. Assuming that the expected trajectory is given It is differentiable on the time interval t∈[0,T), so the purpose of iterative learning control is to design a distributed iterative learning control law So that for all t∈[0,T], when k→∞, Converge to in is the system output value of the kth iteration of the controlled object, m is the number of neighboring nodes of node i, and the coefficient matrices A, B, M, and C i , D i All have corresponding dimensions.

[0028] For linear discrete systems, the design of distributed filters does not consider the input of the system. The output caused by the external disturbance is the deviation of the controlled system output, and the mathematical model of the controlled system is reconstructed:

[0029]

[0030] In the present invention, each control system i adopts a full-order filter of the following form:

[0031]

[0032] in is the filter i’s estimate of the system state, is the estimate of z(k) by filter i. The matrix K ij , is the gain of filter i.

[0033] Then the iterative learning control based on distributed filter is:

[0034]

[0035] First, the gains of the distributed filters are solved.

[0036] For the sake of convenience, consider n filter nodes and define the estimated error vector at time k: System state vector Output estimated error vector and the system parameter vector

[0037]

[0038] definition

[0039]

[0040] Obviously, if but and is a sparse matrix and can be expressed as

[0041]

[0042] in

[0043]

[0044] Given γ>0, if there exists a positive definite matrix R, Q=diag{Q1,Q2,…,Q n}, and matrices X, Y, and Make

[0045]

[0046]

[0047] If it holds, the gain of the filter is

[0048]

[0049] At the same time, the designed distributed filter is exponentially stable and has a given l2-l ∞ Performance indicators.

[0050] Then, the distributed iterative learning law is solved.

[0051] Assume that the dynamic process of the controlled object is as shown in formula (1), given the desired trajectory that can be reached If the following conditions are met, then the distributed learning law (4) makes the output trajectory of the controlled system i converge to the desired trajectory.

[0052]

[0053] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A target trajectory tracking method based on iterative learning of distributed filters, characterized in that : Includes the following steps: Step 1: Given the system parameters of the controlled object, the distributed filter gain is calculated by and Perform offline calculations; Considering n filter nodes, the estimated error vector at time k is defined System state vector Output estimated error vector and the system parameter vector definition when Then a ij =0, and is a sparse matrix, represented as in Given γ>0, if there exists a positive definite matrix R, Q=diag{Q1,Q2,...,Q n }, and matrices X, Y, and The gain of the filter is Step 2: Given the desired tracking trajectory signal of each controlled system, calculate Consider the following discrete-time linear time-invariant controlled system in: is an unknown system state, is the system interference, k is the number of iterations, for all t∈[0,T], when k→∞, Converge to in is the system output value of the controlled object at the kth iteration, m is the number of neighboring nodes of node i, and the coefficient matrix A, B, M, C i ,D i All have corresponding dimensions; Step 3: Calculate the deviation between the output signal of the controlled system and the signal expected to be tracked Step 4: Select the Satisfaction Suitable parameters for And set the number of iterations k; Step 5: Calculate the state estimate of the controlled object at the kth iteration, which is based on the distributed filter and simultaneously fuses the state estimates of its neighboring nodes. calculate calculate Step 6: Information Fusion Step 7: Calculate the output of the controlled object at the kth iteration