Adaptive Control Using Multiple Models and Model Weighting

Abhilash Rajagopal; K. Krishnamurthy

doi:10.1115/IMECE1996-0413

Adaptive Control Using Multiple Models and Model Weighting

Abhilash Rajagopal
, K. Krishnamurthy

Mechanical and Aerospace Engineering

Research output: Contribution to conference › Paper › peer-review

Abstract

A general methodology for the identification and control of dynamical systems with several operating environments and possessing a high degree of uncertainty is presented. Neural networks are used to create multiple models to capture the dynamics of the various environments of the system. Control is effected by combining these models by using an evolutionary strategy. The methodology is applied to the problem of controlling a two-link robotic manipulator in the presence of disturbances and varying load conditions. Simulated results presented show that the proposed methodology yields better results compared to the ones obtained by using a single model or by using multiple models but switching to and tuning the model with the smallest tracking error.

Original language	American English
Pages	805-812
Number of pages	8
DOIs	https://doi.org/10.1115/IMECE1996-0413
State	Published - Nov 1 1996
Event	Proceedings of the 1996 ASME International Mechanical Engineering Congress and Exposition - Atlanta, GA, USA Duration: Nov 17 1996 → Nov 22 1996

Conference

Conference	Proceedings of the 1996 ASME International Mechanical Engineering Congress and Exposition
City	Atlanta, GA, USA
Period	11/17/96 → 11/22/96

ASJC Scopus Subject Areas

Software
Mechanical Engineering

Keywords

Adaptive control systems
Backpropagation
Computer simulation
Genetic algorithms
Identification (control systems)
Learning systems
Manipulators
Mathematical models
Neural networks
Nonlinear control systems
Optimization
Robotics
Standard backpropagation method
Two link robotic manipulator

Disciplines

Mechanical Engineering

Access to Document

10.1115/IMECE1996-0413

Cite this

@conference{f63edc6f21c2475b9c2d5b85f4e85cbb,

title = "Adaptive Control Using Multiple Models and Model Weighting",

abstract = " A general methodology for the identification and control of dynamical systems with several operating environments and possessing a high degree of uncertainty is presented. Neural networks are used to create multiple models to capture the dynamics of the various environments of the system. Control is effected by combining these models by using an evolutionary strategy. The methodology is applied to the problem of controlling a two-link robotic manipulator in the presence of disturbances and varying load conditions. Simulated results presented show that the proposed methodology yields better results compared to the ones obtained by using a single model or by using multiple models but switching to and tuning the model with the smallest tracking error.",

keywords = "Adaptive control systems, Backpropagation, Computer simulation, Genetic algorithms, Identification (control systems), Learning systems, Manipulators, Mathematical models, Neural networks, Nonlinear control systems, Optimization, Robotics, Standard backpropagation method, Two link robotic manipulator",

author = "Abhilash Rajagopal and K. Krishnamurthy",

note = "Publisher Copyright: {\textcopyright} 1996 American Society of Mechanical Engineers (ASME). All rights reserved.; Proceedings of the 1996 ASME International Mechanical Engineering Congress and Exposition ; Conference date: 17-11-1996 Through 22-11-1996",

year = "1996",

month = nov,

day = "1",

doi = "10.1115/IMECE1996-0413",

language = "American English",

pages = "805--812",

}

TY - CONF

T1 - Adaptive Control Using Multiple Models and Model Weighting

AU - Rajagopal, Abhilash

AU - Krishnamurthy, K.

PY - 1996/11/1

Y1 - 1996/11/1

N2 - A general methodology for the identification and control of dynamical systems with several operating environments and possessing a high degree of uncertainty is presented. Neural networks are used to create multiple models to capture the dynamics of the various environments of the system. Control is effected by combining these models by using an evolutionary strategy. The methodology is applied to the problem of controlling a two-link robotic manipulator in the presence of disturbances and varying load conditions. Simulated results presented show that the proposed methodology yields better results compared to the ones obtained by using a single model or by using multiple models but switching to and tuning the model with the smallest tracking error.

AB - A general methodology for the identification and control of dynamical systems with several operating environments and possessing a high degree of uncertainty is presented. Neural networks are used to create multiple models to capture the dynamics of the various environments of the system. Control is effected by combining these models by using an evolutionary strategy. The methodology is applied to the problem of controlling a two-link robotic manipulator in the presence of disturbances and varying load conditions. Simulated results presented show that the proposed methodology yields better results compared to the ones obtained by using a single model or by using multiple models but switching to and tuning the model with the smallest tracking error.

KW - Adaptive control systems

KW - Backpropagation

KW - Computer simulation

KW - Genetic algorithms

KW - Identification (control systems)

KW - Learning systems

KW - Manipulators

KW - Mathematical models

KW - Neural networks

KW - Nonlinear control systems

KW - Optimization

KW - Robotics

KW - Standard backpropagation method

KW - Two link robotic manipulator

UR - https://www.scopus.com/pages/publications/0030409167

UR - https://www.scopus.com/pages/publications/0030409167#tab=citedBy

U2 - 10.1115/IMECE1996-0413

DO - 10.1115/IMECE1996-0413

M3 - Paper

SP - 805

EP - 812

T2 - Proceedings of the 1996 ASME International Mechanical Engineering Congress and Exposition

Y2 - 17 November 1996 through 22 November 1996

ER -

Adaptive Control Using Multiple Models and Model Weighting

Abstract

Conference

ASJC Scopus Subject Areas

Keywords

Disciplines

Access to Document

Other files and links

Fingerprint

Cite this