A new evolution strategy and its application to solving optimal control problems

M. M A Hashem, Keigo Watanabe, Kiyotaka Izumi

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Evolution strategies (ESs) are search algorithms which imitate the principles of natural evolution as a method to solve parameter optimization problems numerically. The effectiveness and simplicity of ES algorithms have lead many people to believe that they are the methods of choice for difficult, real-life problems superseding traditional search techniques. However, the inherent strength of the ES algorithms largely depends upon the choice of a suitable crossover and mutation technique in their application domains. This paper discusses a new ES in which both a subpopulation-based arithmetical crossover (SBAC) and a time-variant mutation (TVM) operator are used. The SBAC operator explores promising areas in the search space with different directivity while the TVM operator exploits fast (but not premature) convergence with high precision results. Thus, a balance between exploration and exploitation is achieved in the evolutionary process with these combined efforts. The TVM also acts as a fine local tuner at the converging stages for high precision solutions. Its action depends upon the age of the populations, and its performance is quite different from the Uniform Mutation (UM) operation. The efficacy of the proposed methods is illustrated by solving discrete-time optimal control models which are frequently used in the applications.

Original languageEnglish
Pages (from-to)406-412
Number of pages7
JournalJSME International Journal, Series C: Dynamics, Control, Robotics, Design and Menufacturing
Volume41
Issue number3
Publication statusPublished - Sep 1998
Externally publishedYes

Keywords

  • Arithmetical Crossover
  • Discrete-Time Optimal Control
  • Evolution Strategy
  • Evolutionary Computation
  • Intelligent Systems
  • Linear-Quadratic Control
  • Mutation
  • Push-Cart Control

ASJC Scopus subject areas

  • Engineering(all)
  • Industrial and Manufacturing Engineering
  • Mechanical Engineering

Cite this

A new evolution strategy and its application to solving optimal control problems. / Hashem, M. M A; Watanabe, Keigo; Izumi, Kiyotaka.

In: JSME International Journal, Series C: Dynamics, Control, Robotics, Design and Menufacturing, Vol. 41, No. 3, 09.1998, p. 406-412.

Research output: Contribution to journalArticle

@article{1074a17bb88b4d3aaa08e96485fae462,
title = "A new evolution strategy and its application to solving optimal control problems",
abstract = "Evolution strategies (ESs) are search algorithms which imitate the principles of natural evolution as a method to solve parameter optimization problems numerically. The effectiveness and simplicity of ES algorithms have lead many people to believe that they are the methods of choice for difficult, real-life problems superseding traditional search techniques. However, the inherent strength of the ES algorithms largely depends upon the choice of a suitable crossover and mutation technique in their application domains. This paper discusses a new ES in which both a subpopulation-based arithmetical crossover (SBAC) and a time-variant mutation (TVM) operator are used. The SBAC operator explores promising areas in the search space with different directivity while the TVM operator exploits fast (but not premature) convergence with high precision results. Thus, a balance between exploration and exploitation is achieved in the evolutionary process with these combined efforts. The TVM also acts as a fine local tuner at the converging stages for high precision solutions. Its action depends upon the age of the populations, and its performance is quite different from the Uniform Mutation (UM) operation. The efficacy of the proposed methods is illustrated by solving discrete-time optimal control models which are frequently used in the applications.",
keywords = "Arithmetical Crossover, Discrete-Time Optimal Control, Evolution Strategy, Evolutionary Computation, Intelligent Systems, Linear-Quadratic Control, Mutation, Push-Cart Control",
author = "Hashem, {M. M A} and Keigo Watanabe and Kiyotaka Izumi",
year = "1998",
month = "9",
language = "English",
volume = "41",
pages = "406--412",
journal = "JSME International Journal, Series 3: Vibration, Control Engineering, Engineering for Industry",
issn = "0914-8825",
publisher = "Japan Society of Mechanical Engineers",
number = "3",

}

TY - JOUR

T1 - A new evolution strategy and its application to solving optimal control problems

AU - Hashem, M. M A

AU - Watanabe, Keigo

AU - Izumi, Kiyotaka

PY - 1998/9

Y1 - 1998/9

N2 - Evolution strategies (ESs) are search algorithms which imitate the principles of natural evolution as a method to solve parameter optimization problems numerically. The effectiveness and simplicity of ES algorithms have lead many people to believe that they are the methods of choice for difficult, real-life problems superseding traditional search techniques. However, the inherent strength of the ES algorithms largely depends upon the choice of a suitable crossover and mutation technique in their application domains. This paper discusses a new ES in which both a subpopulation-based arithmetical crossover (SBAC) and a time-variant mutation (TVM) operator are used. The SBAC operator explores promising areas in the search space with different directivity while the TVM operator exploits fast (but not premature) convergence with high precision results. Thus, a balance between exploration and exploitation is achieved in the evolutionary process with these combined efforts. The TVM also acts as a fine local tuner at the converging stages for high precision solutions. Its action depends upon the age of the populations, and its performance is quite different from the Uniform Mutation (UM) operation. The efficacy of the proposed methods is illustrated by solving discrete-time optimal control models which are frequently used in the applications.

AB - Evolution strategies (ESs) are search algorithms which imitate the principles of natural evolution as a method to solve parameter optimization problems numerically. The effectiveness and simplicity of ES algorithms have lead many people to believe that they are the methods of choice for difficult, real-life problems superseding traditional search techniques. However, the inherent strength of the ES algorithms largely depends upon the choice of a suitable crossover and mutation technique in their application domains. This paper discusses a new ES in which both a subpopulation-based arithmetical crossover (SBAC) and a time-variant mutation (TVM) operator are used. The SBAC operator explores promising areas in the search space with different directivity while the TVM operator exploits fast (but not premature) convergence with high precision results. Thus, a balance between exploration and exploitation is achieved in the evolutionary process with these combined efforts. The TVM also acts as a fine local tuner at the converging stages for high precision solutions. Its action depends upon the age of the populations, and its performance is quite different from the Uniform Mutation (UM) operation. The efficacy of the proposed methods is illustrated by solving discrete-time optimal control models which are frequently used in the applications.

KW - Arithmetical Crossover

KW - Discrete-Time Optimal Control

KW - Evolution Strategy

KW - Evolutionary Computation

KW - Intelligent Systems

KW - Linear-Quadratic Control

KW - Mutation

KW - Push-Cart Control

UR - http://www.scopus.com/inward/record.url?scp=11744332007&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=11744332007&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:11744332007

VL - 41

SP - 406

EP - 412

JO - JSME International Journal, Series 3: Vibration, Control Engineering, Engineering for Industry

JF - JSME International Journal, Series 3: Vibration, Control Engineering, Engineering for Industry

SN - 0914-8825

IS - 3

ER -