Learning identification and control for repetitive linear time-varying systems

Liu, Nanjun

Permalink

https://hdl.handle.net/2142/50586 Copy

Description

Title

Learning identification and control for repetitive linear time-varying systems

Author(s)

Liu, Nanjun

Issue Date

2014-09-16

Director of Research (if dissertation) or Advisor (if thesis)

Alleyne, Andrew G.

Doctoral Committee Chair(s)

Alleyne, Andrew G.

Committee Member(s)

• Beck, Carolyn L.
• Sreenivas, Ramavarapu S.
• Salapaka, Srinivasa M.

Department of Study

Mechanical Sci & Engineering

Discipline

Mechanical Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• System Identification
• Linear Systems
• Iterative Learning Control

Abstract

There are many manufacturing systems that can be described as Linear Time-Varying (LTV) systems that have large parameter variations. When using a feedback controller for these systems, the lag in transient tracking response is always inevitable. The high-level precision tracking requirements of these systems provide new challenges in algorithm development. In this research, we consider identification and precise motion control for repetitive LTV systems. In particular, we focus on the iterative learning concept, which capitalizes on the repetition of task to update and improve identification and control with each trial. This concept is originally developed from Iterative Learning Control (ILC), which reduces the tracking error of the current iteration by incorporating information learnt from previous executions. In this research, we explore the extension of the ILC concept to both identification and control. This dissertation develops two contributions to the identification and control of repetitive LTV systems. First, an Iterative Learning Identification (ILI) algorithm is developed for identifying the parameters of repetitive LTV systems. The proposed ILI scheme takes advantage of the repetitive nature of the system, and non-causal data is used to minimize the estimation transient. The design, analysis, simulation and experimental results for ILI on LTV systems are presented in the thesis. Second, we integrate the identification with norm-optimal ILC design approach. These techniques are used to improve the convergence speed of norm-optimal ILC when the LTV model of the system is not initially available. The integrated ILI and ILC is applied to a pick and place robot with a time-varying mass and yields an improved convergence speed over an ILC controller developed from a recursive model.

Graduation Semester

2014-08

Permalink

http://hdl.handle.net/2142/50586

Copyright and License Information

Copyright 2014 Nanjun liu

Owning Collections