Citation
Andrew, A.M. (2002), "Soft Computing: New Trends and Applications", Kybernetes, Vol. 31 No. 3/4. https://doi.org/10.1108/k.2002.06731cae.002
Publisher
:Emerald Group Publishing Limited
Copyright © 2002, MCB UP Limited
Soft Computing: New Trends and Applications
by L. Fortuna, G. Rizzotto, M. Lavorgna, G. Nunnari, M.G. Xibilia and R. CaponettoSpringerLondon2001xii + 257 pp.ISBN 1-85233-308-1, ISSN 1439-2232paperback, with CD ROM, £29.50 or $49.95Advanced Textbooks in Control and Signal Processing Series
The term "soft computing" is attributed to Lotfi Zadeh and indicates a range of techniques that produce useful results without the rigorous analysis that is a feature of traditional methods. Specific mention is made of fuzzy logic and artificial neural nets, as well as the optimisation techniques of evolutionary algorithms and simulated annealing. It is acknowledged that there are numerous conference proceedings and other publications that treat these topics separately, but here a selection of the most useful is covered in one volume. This is valuable because important applications have been made that combine methods so as to produce systems that can be described as neuro-fuzzy, fuzzy- genetic or as using fuzzy cellular neural networks. The emphasis is firmly on applications and the treatment is driven by the needs of industry rather than by other considerations such as speculation about the working of real nervous systems.
All of the topics are developed from, essentially, first principles, and a number of quite advanced applications are described. To cover all this in a fairly small book it is necessary that the treatment is in places somewhat terse though always chatty and readable. Each chapter is followed by a set of references that will give useful support. The references are mainly to modern works and do not necessarily honour the pioneers, so that there is, for example, no mention of McCulloch and Pitts. Attention is necessarily focused on those developments that have proved to have practical utility, and the treatment of neural nets is largely concerned with multi-layer perceptrons, with an acknowledgement of Kohonen nets but no mention of the Hopfield variety.
The treatment of fuzzy logic, and then of fuzzy control, in the first two chapters after the Introduction, is particularly welcome since it spells out the application of such control in an application of a suitable degree of complexity, namely a device containing a switching transistor, as well as an inductor, capacitor and rectifier, that can be used for DC voltage adjustment.
Cellular neural nets also receive extensive treatment. These are regular arrays of neuron-like elements, each with connections only to its immediate neighbours. An example is the game of "Life". Such arrays have important applications in image processing, and their analysis requires the invoking of Chaos Theory. This is developed in the book in considerable detail, introducing the ideas of chaotic dynamics and strange attractors. The development is not mathematically painless, but is probably as near to it as is possible, and the book is worth consulting for this feature alone.
The applications mentioned range widely. One chapter is devoted to description of a robot for picking citrus fruit, where the complex dynamics of the robotic arm were allowed for using neural nets in the controller. The system is still experimental and some alternative arrangements are discussed, with no data about actual performance in an orchard, but even so it is a development that will arouse widespread interest.
The book comes with a CD ROM, which seems to have been an afterthought as there is no mention of it in the book. The author of the CD ROM introduces himself as Francesco Zivillica. On it there is an assortment of additional pieces of theory along with animated demonstrations, a number of them interactive. The demonstrations include the formation of swirling spiral patterns that are important attractor states of cellular neural nets, as well as some of the remarkable effects achievable in the game of "Life". There is also a presentation of a means of modelling by use of a fuzzy-genetic algorithm that appears to be additional to the material presented in the book itself, and animated views of some mobile robots.
The book, with its CD ROM, is a valuable addition to the literature. It is suitable as a teaching text, and will certainly facilitate practical exploitation of combinations of the different kinds of soft computing. The treatments of such topics as fuzzy control and chaos theory will be welcomed as instructive texts by those who find it easiest to get to grips with abstruse theories in specific practical contexts.
Alex M. Andrew