A PID-type iterative learning control method based on neural network

A technology of iterative learning control and neural network, applied in the field of PID iterative learning control based on neural network, to achieve good precision and efficient iterative learning and tracking

Active Publication Date: 2022-05-10
HARBIN INST OF TECH SHENZHEN GRADUATE SCHOOL
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, there are not many research results on iterative learning problems of general non-affine nonlinear systems or nonlinear systems with unknown exact models.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A PID-type iterative learning control method based on neural network
  • A PID-type iterative learning control method based on neural network
  • A PID-type iterative learning control method based on neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0079] Consider the following system:

[0080]

[0081] Obviously, this is a non-affine nonlinear system. The expected output trajectory in the iterative learning task is:

[0082] the y d (t)=0.5×(-1) round(t / 10) t∈I[1,100]

[0083] We adopt the iterative learning scheme formula (17) proposed in the present invention to control the system, and now we pre-set the structure of the neural network and the initial values ​​of the parameters. The three neural networks used to estimate the gain of the PID controller all use a hidden layer with three nodes, and the inputs of the three neural networks are:

[0084]

[0085] For all i∈I[1,+∞), t∈I[1,T], we set the radial basis function center of the neural network to be [0 0 0] T , with a width of 100; for all t∈I[1,T], the input of the first iterative neural network is set to z S,1 (t)=[0 0 0] T . The weight of the first iteration is w P,1 (t) = [0.2 0 0] T ,w I,1 (t)=[0 0 0] T ,w D,1 (t) = [0 00] T ; For all t∈I[...

Embodiment 2

[0088] Such as figure 2 is a pick-and-place robot model (refer to non-patent literature 2: Liu N. Learning identification and control for repetitive linear time-varying systems [D]. University of Illinois at Urbana-Champaign, 2014.), for the rotation angle θ of the manipulator, we get the state variable x 1 = θ and And taking the angular velocity as the system output, the following second-order linear mathematical model can be obtained:

[0089]

[0090] Where β=2Nm / rad is the viscous friction coefficient, K t =100 is the gain of the actuator, M(t) is the mass at the bottom of the time-varying manipulator, and L=0.1m is the length of the manipulator. During the pick-and-place process of the robotic arm, the heavy object is grasped in the 5th second and put down in the 10th second. The mass of M(t) also changes from 1kg to 10kg and then back to 1kg.

[0091] Take the sampling time T s =0.01s, the discretization model (19) in the time interval [1,10] can obtain the fo...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention proposes a neural network-based PID type iterative learning control method, which belongs to the field of control science and engineering. For the general nonlinear difference system, the author first gives an iterative learning PID controller, and then uses three RBF neural networks to fit the dynamic parameters in the PID controller. According to the gradient descent method, the weight of the neural network is estimated and the expression of updating the weight of the neural network is obtained. Since the precise model of the controlled system is complex or unknown, we cannot obtain the true value of the system pseudo partial derivative (PPD) in the neural network weight estimation expression, and we use its estimated value as a numerical substitute. This produces a non-linear iterative learning control algorithm that can be run. Finally, examples illustrate that the algorithm is effective in numerical simulations of nonlinear systems and pick-and-place robot systems. Therefore, this algorithm can be applied to general (unknown) nonlinear difference systems.

Description

technical field [0001] The invention relates to the field of control science and engineering, in particular to a neural network-based PID iterative learning control method. Background technique [0002] Iterative learning is a control method in which the controlled system continuously absorbs previous experience for repeated learning in a limited time interval. It can be understood as: when the controlled system performs the i-th operation, the controller uses the input (output) data and errors of the (i-1) and previous times to make the controlled system Get a better execution effect. According to this method, the controller is constantly corrected, so that the controlled system moves towards the desired trajectory. [0003] In the past three decades, iterative learning control has achieved rich results both in theory and in application. However, these efforts mainly focus on solving iterative learning control problems for linear or affine nonlinear systems of known mode...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Patents(China)
IPC IPC(8): G05B11/42
CPCG05B11/42
Inventor 常明方吴爱国
Owner HARBIN INST OF TECH SHENZHEN GRADUATE SCHOOL
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products