|
Fuzzy Evolutionary Computation
Witold Pedrycz, Editor
Kluwer Academic Publishers, Boston, London, Dordrecht, 1997
ISBN 0-7923-9942-0
With this excellent book on fuzzy evolutionary computation the editor has performed a
remarkable threefold service. First, the book is very informative as it represents an up-to-date
view of the rapidly growing field of interdisciplinary research and, second, it is a very
successful pedagogical example to bring together innovative practical and advanced issues. Third,
most of the chapters deal with exercises or open problems which are of special interest to the
reader.
Let us survey, very briefly, the contents of the reviewed book. The book includes 13 chapters
and a subject index, which are structured into three parts. Part 1 deals with fundamentals of fuzzy
evolutionary computation and Part 2 with the methodology and algorithms, whereby each chapter
contains relevant references. Additionally, Part 3 introduces an indexed bibliography of genetic
algorithms with fuzzy logic, which contains references to scientific works published in journals,
conference proceedings and books with the theme genetic algorithms and fuzzy techniques. To gain a
general impression about the material covered by this edited volume, we will go very briefly over
the chapters. The first Chapter 1.1, authored by Z. Michalewicz et al., is addressed to
evolutionary algorithms and gives some background and emphasis on handling data structures and
genetic operations. The presented material is definitely highly informative and guides the reader
through the main paradigms of evolutionary algorithms. Indeed the essentials of the authors
"why evolutionary algorithms perform well: (1) independent sampling is provided by populations
of candidate solutions, (2) selection is a mechanism that preserves good solutions, and (3) partial
solutions can be efficiently modified and combined through various genetic operators" are true
and stimulating. Chapter 1.2, authored by O. Cordon et al., deals with a very short sketch of some
statements about the combination of fuzzy logic and evolutionary computation with often not
representative references. The presentation gives an imagination how fuzzy rule based systems can
be an important tool for modeling complex systems. The fuzzy genetic system techniques in Chapter
1.3 proposed by M.A. Lee and H. Esbensen deal with genetic approaches to adapting fuzzy knowledge
structures and fuzzy knowledge structures for adapting genetic search techniques. Key ingredient of
both is a multiobjective optimization of a genetic algorithm. A meta-level evolutionary algorithm
technique for automatically designing search control strategies for multiobjective evolutionary
algorithms is described for intelligent systems design tools and components.
The problem of hierarchical trajectory planning for robot manipulators within the framework
of genetic algorithms is studied by T. Fukuda et al. (Chapter 2.1). To that purpose a
virus-evolutionary genetic algorithm is proposed and a virus operator is introduced into the
genetic algorithm as new searching operators. Some applications examples to mobile robots,
redundant manipulators and biped locomotion robots are presented. The Chapter 2.2 by T. Furuhashi
describes a new coding method, DNA coding method, and a mechanism of the development from the
artificial DNA or artificial amino acids. Each artificial amino acid can have several meanings: An
amino acid can be translated as an input variable or a form of membership function. A sequence of
amino acids can represent a fuzzy rule, and moreover, sets of fuzzy rules by a DNA chromosome. The
given application for controlling a mobile robot demonstrates its applicability. The Chapter 2.3 by
H. Ishibuchi et al. deals with a fuzzy rule-based classification system for both classification and
rule selection. The classification results can be obtained by the proposed fuzzy reasoning method
and the genetic-algorithm-based rule selection is described with two objectives: to minimize the
number of fuzzy if-then rules and to maximize the number of correctly classified patterns. Computer
simulation results on iris and wine data demonstrate the applicability of the novel approach. The
paper by M. Sakawa and T. Shibano (Chapter 2.4) proposes an approach of genetic algorithms in a
more biological sense. The aim of this concept takes into account the representation of the
individuals in a double string order. Moreover new genetic operators are introduced and differences
between conventional GAs and the proposed GAs are indicated. In this context an interactive fuzzy
satisfying method for multiobjective multidimensional 0-1 knapsack problems is proposed and
numerical examples demonstrate both feasibility and effectiveness of the method. Chapter 2.5,
authored by J. Kacprzyk, is devoted to the use of genetic algorithms for the solution of multistage
optimal control of fuzzy dynamical systems. An optimal sequence of controls is sought which
maximizes the fuzzy decision over a fixed and specified planning horizon. It can be shown that the
genetic algorithm approach proposed is conceptually simpler than traditionally employed techniques,
e.g. dynamic programming and branch-and-bound. The problem of learning using genetic algorithms in
neural fuzzy control systems is studied by D.A. Linkens and H.O. Nyongesa (Chapter 2.6). The goal
is to construct a system which combines neural and GA learning in different but complementary
learning methods together with fuzzy logic. The fusion of all three methodologies in a learning
control context is demonstrated on an aero-engine control. The Chapter 2.7, authored by G.
Vulkovich and J.Y. Lee, presents a dynamic fuzzy logic system structure and an indirect adaptive
control approach based on this structure is developed. Stability properties are investigated on
this novel approach and applications of nonlinear systems illustrate the effectiveness and
reasonable good system performance. The objective of Chapter 2.8, authored by L. Magdalena and J.R.
Velasco, deals with some general ideas of learning needed for the construction of an evolutionary
fuzzy controller. Finally, Chapter 2.9 by O. Nelles gives a remarkable survey about the generation
of fuzzy rules by genetic algorithms, optimization and combination of GAs with classical
optimization techniques in detail. A real world example demonstrates its applicability.
The book provides an excellent research-oriented overview of the subject with extensive
references to the published literature. In all chapters, except 1.2, the fundamental theory and
important algorithms are presented in an appealing pedagogical profile. Chapter 1.2 deals with
possible applications and has a more catalogue-based style. To unify the style of the book all
chapters should have an abstract or not.
The applications covered in the book are definitely thought-provoking and highly innovative.
New ideas are sketched, moreover, the chapters include characteristic applications and describe
real-world implementations.
Overall, the balance of the weaknesses and the strengths of this book is definitely positive,
it can be seen as a genuine challenge of fuzzy evolutionary computation which is a pleasure to
read.
Professor Dr. A. Grauel
Department of Mathematical Science
University of Paderborn, Campus Soest |
Book Review