Cited time in webofscience Cited time in scopus

Data-driven Motion Control Framework of Model-based Approach for Multi-axis Flexible System

Title
Data-driven Motion Control Framework of Model-based Approach for Multi-axis Flexible System
Author(s)
Hanul Jung
DGIST Authors
Hanul JungSehoon OhChang-Wan Ha
Advisor
오세훈
Co-Advisor(s)
Chang-Wan Ha
Issued Date
2024
Awarded Date
2024-02-01
Type
Thesis
Description
Data-driven optimization;Model-based control;System identification;Order selection;Disturbance observer;Convex optimization;MIMO flexible systems
Abstract
This thesis introduces a novel data-driven motion control framework that synergistically integrates with a model-based approach, specifically tailored for the control of multi-axis flex- ible systems. Traditional model-based control strategies, while robust, often require precise mathematical models that can be computationally intensive and difficult to obtain for complex systems. Conversely, purely data-driven methods, though adaptive and less reliant on model accuracy, can lack the theoretical robustness of model-based approaches. This work bridges this gap by developing an innovative algorithm that leverages the strengths of both paradigms. This means leveraging measured data to improve and optimize model-based controllers. The proposed framework rests on an iterative algorithm that uses measured data to converge on optimal control parameters, minimizing reliance on a priori system models and accommodating system non-linearities and uncertainties. This methodology is based on a rigorous theoretical foundation encompassing system identification techniques, optimization algorithms, and stability analysis to ensure that the con- trol framework not only adapts to the inherent flexibility and dynamics of the system, but also complies with performance criteria. Extensive simulation and experimental studies are presented that demonstrate the effec- tiveness of the data-driven control framework in managing the complexity of multi-axis flexible systems. Results indicate significant improvements in system performance through improved robustness to model inconsistencies, improved suppression of unmodeled dynamics and exter- nal disturbances. In summary, this paper contributes to the control engineering domain by presenting a con- trol framework that optimizes model-based control strategies through data-driven methods. The implications of this study are far-reaching and provide a versatile control architecture that can be extended to a variety of multi-axis flexible systems within a variety of industrial applications. Keywords: Data-driven optimization, Model-based control, System identification, Order se- lection, Disturbance observer, Convex optimization, MIMO flexible systems|이 논문은 특히 다축 유연한 시스템의 제어를 위해 맞춤화된 모델 기반 접근 방식과 시너지 효과적으로 통합되는 새로운 데이터 기반 모션 제어 프레임워크를 소개합니다. 전통적인 모델 기반 제어 전략은 강력하기는 하지만 계산 집약적이고 복잡한 시스템에서는 얻기 어려운 정밀한 수학적 모델이 필요한 경우가 많습니다. 반대로, 순전히 데이터 기반 방법은 적응성이 뛰어나고 모델 정확도에 덜 의존하지만 모델 기반 접근 방식의 이론적 견고성이 부족할 수 있습니다.

이 작업은 두 패러다임의 장점을 활용하는 혁신적인 알고리즘을 개발하여 이러한 격차를 해소합니다. 이는 측정된 데이터를 활용하여 모델 기반 컨트롤러를 개선하고 최적화하는 것을 의미합니다. 제안된 프레임워크는 측정된 데이터를 사용하여 최적의 제어 매개변수에 수렴하고 선험적 시스템 모델에 대한 의존도를 최소화하며 시스템 비선형성 및 불확실성을 수용하는 반복 알고리즘을 기반으로 합니다.

이 방법론은 제어 프레임워크가 시스템의 고유한 유연성과 역동성에 적응할 뿐만 아니라 성능 기준도 준수하도록 보장하기 위해 시스템 식별 기술, 최적화 알고리즘 및 안정성 분석을 포괄하는 엄격한 이론적 기반을 기반으로 합니다.

다축 유연한 시스템의 복잡성을 관리하는 데 있어 데이터 기반 제어 프레임워크의 효율성을 입증하는 광범위한 시뮬레이션 및 실험 연구가 제시됩니다. 결과는 모델 불일치에 대한 견고성 향상, 모델링되지 않은 역학 및 외부 방해에 대한 억제 향상을 통해 시스템 성능이 크게 향상되었음을 나타냅니다.

요약하면, 본 논문은 데이터 기반 방법을 통해 모델 기반 제어 전략을 최적화하는 제어 프레임워크를 제시함으로써 제어 엔지니어링 영역에 기여합니다. 이 연구의 의미는 광범위하며 다양한 산업 응용 분야 내에서 다양한 다축 유연한 시스템으로 확장할 수 있는 다목적 제어 아키텍처를 제공합니다.
Table Of Contents
List of Contents
Abstract i
List of Contents . ii
List of Tables vii
List of Figures viii
I.INTRODUCTION . 1
1.1 RESEARCH BACKGROUND 1
1.2 LITERATURE REVIEW AND PROBLEM STATEMENTS . 2
1.2.1 Decoupling Control for Dual-Drive Gantry Stage 2
1.2.2 Parametric System Identification Using Frequency Response Function . 4
1.2.3 Data-Driven Optimization for Model-based Controller 7
1.3 CONTRIBUTION POINTS OF THESIS 9
1.4 OUTLINE OF THESIS 12
II.MULTI-INPUT-MULTI-OUTPUT SYSTEM DECOMPOSITION
FROM FREQUENCY RESPONSE PERSPECTIVE . 14
2.1 INTRODUCTION 14
2.2 PROBLEM FORMULATION: ANALYSIS OF COUPLING DYNAMICS 16
2.2.1 Dynamics Analysis of the Dual-drive Gantry Stage 16
2.2.2 Mechanical Coupling in Conventional Actuator-based Dynamics . 18
2.2.3 Effect of Coupling in Actuator-based Dynamics . 20
– ii –
2.3 DATA-DRIVEN MODE DECOMPOSITION AND RECONFIGURED DY-
NAMICS FROM PERSPECTIVE OF TASK-BASED MOTIONS . 23
2.3.1 Canonical Polyadic Decomposition-based Dynamics Analysis 24
2.3.2 Reconfiguration of Dual-drive Gantry System Dynamics based on CPD
and Task Motion 27
2.4 DESIGN AND ANALYSIS OF TWO-DEGREE-OF-FREEDOM CASCADE
CONTROL BASED ON TASK-BASED DYNAMICS . 31
2.4.1 Application of Task-based Dynamics for Model-based Cascade Control 32
2.4.2 Improvement of Robust Stability of the Model-based Control using
Task-based Dynamics 35
2.5 EXPERIMENTAL RESULTS 37
2.5.1 Experimental Set-up 37
2.5.2 Improvement of Bandwidth by the Proposed Controller 37
2.5.3 Robust Performance Improvement of Task-based Disturbance Observer
against Additional Mass . 40
2.5.4 Decoupling Effect by the Proposed Mode Decomposition and Control . 42
2.6 CONCLUSION . 44
III.SYSTEM PARAMETRIC IDENTIFICATION BASED ON RO-
BUST STABILITY CRITERION 45
3.1 INTRODUCTION 45
3.2 ANALYSIS OF DISTURBANCE OBSERVER AND OPTIMIZATION CHAR-
ACTERISTICS . 47
3.2.1 Design of Disturbance Observer 48
– iii –
3.2.2 Robust Stability Margin Analysis of Disturbance Observer . 50
3.2.3 Cost Function Design and Its Convexity 51
3.3 DISTURBANCE OBSERVER RELEVANT PARAMETRIC SYSTEM IDEN-
TIFICATION 53
3.3.1 Brief of Weighted Iterative Least Square Algorithm using Modeling Error 53
3.3.2 DOB-Relevant System Identification . 55
3.3.3 ℓ2-norm approximation for ℓ∞-norm?? 57
3.3.4 Procedure of the DOB-relevant system identification algorithm 58
3.4 EXPERIMENTAL VERIFICATION 60
3.4.1 Condition and Setting 60
3.4.2 Performance Validation Compared to the Conventional Method . 61
3.4.3 Characteristics of DOB-relevant Identification According to Desired
Bandwidth 66
3.5 CONCLUSION AND FUTURE WORKS 70
IV.MODEL ORDER SELECTION RELATED TO ROBUST STABIL-
ITY CRITERION . 71
4.1 INTRODUCTION 72
4.2 PROBLEM FORMULATION 73
4.2.1 Disturbance Observer based on Nominal Model . 73
4.2.2 Parametric Identification with Noisy & Discrete Data and Model Order
Selection 74
4.3 PARAMETRIC IDENTIFICATION BASED ON FREQUENCY RESPONSE
FUNCTION 76
– iv –
4.3.1 System Description based on Frequency Response Model . 76
4.3.2 Parametric Identification using Frequency Response Model . 77
4.4 MODEL ORDER SELECTION BASED ON ROBUST STABILITY 79
4.5 NUMERICAL SIMULATION OF PROPOSED METHOD 81
4.5.1 Simulation Condition and Setting 81
4.5.2 Model order selection based on robust stability criterion 83
4.6 CONCLUSION AND FUTURE WORKS 85
V.DATA-DRIVEN CONTROL OPTIMIZATION FRAMEWORKS FOR
MODEL-BASED CONTROLLER . 86
5.1 INTRODUCTION 86
5.1.1 Backgrounds . 86
5.1.2 Studies on Data-driven Controller Tuning . 88
5.1.3 Novel Data-driven Optimization Algorithm 89
5.2 ANALYSIS OF INTEGRATED CONTROL AND OPTIMIZATION CHAR-
ACTERISTICS . 91
5.2.1 Design of Integrated Control 91
5.2.2 Cost Function Design for Optimization and Its Convexity 93
5.2.3 Robust Stability Analysis for Integrated Control . 95
5.3 PROPOSED DATA-DRIVEN INTEGRATED CONTROLLER OPTIMIZATION100
5.3.1 Iterative Feedback Tuning 99
5.3.2 Gradient Calculation for Integrated Control 101
5.3.3 Unbiased Estimation of Cost Function Gradient . 104
5.3.4 Convergence Analysis 106
– v –
5.4 EXPERIMENTAL VERIFICATION 107
5.4.1 Experimental Set-up 108
5.4.2 Optimization Performance with Different Initial Control Parameters 109
5.4.3 Optimization Performance With Various Plant Conditions . 112
5.5 CONCLUSION . 116
VI.INTEGRATED PROCEDURE OF THESIS RESEARCH 117
6.1 BRIEF OF INTEGRATED PROCEDURE 117
6.2 ENTIRE PROCEDURE OF 118
6.2.1 Frequency response & initial setting . 118
6.2.2 Data-based system decoupling . 120
6.2.3 DOB-relevant system identification with order selection algorithm 122
6.2.4 Data-driven control parameter optimization 124
6.3 CONCLUSION . 126
VII.CONCLUSION AND RECOMMANDATION 127
7.1 CONCLUSIONS 127
VIII.ACHIEVEMENTS 128
8.1 Journal Papers 128
8.2 International Conference Papers . 128
8.3 Domestic Conference Papers 130
8.4 Awards 130
References . 131
– vi –
URI
http://hdl.handle.net/20.500.11750/48012

http://dgist.dcollection.net/common/orgView/200000724305
DOI
10.22677/THESIS.200000724305
Degree
Doctor
Department
Department of Robotics and Mechatronics Engineering
Publisher
DGIST
Related Researcher
  • 오세훈 Oh, Sehoon
  • Research Interests Research on Human-friendly motion control; Development of human assistance;rehabilitation system; Design of robotic system based on human musculoskeletal system; Analysis of human walking dynamics and its application to robotics; 친인간적인 운동제어 설계연구; 인간 보조;재활 시스템의 설계 및 개발연구; 인간 근골격계에 기초한 로봇기구 개발연구; 보행운동 분석과 모델 및 로봇기구에의 응용
Files in This Item:

There are no files associated with this item.

Appears in Collections:
Department of Robotics and Mechatronics Engineering Theses Ph.D.

qrcode

  • twitter
  • facebook
  • mendeley

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE