Information Technology Reference
In-Depth Information
5
Experiences and Obstacles in Industrial
Applications of Intelligent Systems
Leonardo M. Reyneri
1
and Valentina Colla
2
1
Politecnico di Torino, Dipartimento di Elettronica e Telecomunicazioni, Torino,
2
Scuola Superiore Sant'Anna, TeCIP Institute, PERCRO, Pisa,
Italy
1. Introduction
Neural networks
and
fuzzy systems
are well known soft computing techniques, which date
back several decades since the preliminary work of McCulloch and Pitts, Grossberg,
Zadeh, and dozens of other precursors. At first, the neural network was believed to be
"simple and workable solution" for all the difficult problems can be dealt with, and then
gave rise to a broad interest in research around the world and garnered a lot of funding.
During this preliminary period, many theories have been developed, analyzed and
applied.
Later, the domain of neural networks and fuzzy systems has broadened and also many
other algorithms and methods have been collected under the term of
Soft Computing
and,
more general,
Intelligent Systems
. These include, among others, neural networks, fuzzy
logic, wavelet networks, genetic algorithms, expert systems, etc... It was then discovered
that several simple problems (the so-called "toy problems") actually found very simple
solutions using intelligent systems. On the other hand, difficult problems (for example,
handwriting recognition and most problems of industrial relevance), still could not be
completely resolved, even if intelligent systems could contribute to simplify their
solution.
Today, after several decades of alternating interest of the scientific and industrial
community, after the publication of tens of thousands of theoretical and practical papers,
and after several attempts to apply them in a large number of application domains,
intelligent systems are now reaching a rather
mature phase
. People have begun to understand
the real capabilities, potentials, limitations and disadvantages, so they are on the right path
towards a widespread adoption, without excessive and inappropriate enthusiasm, but also,
more importantly, with a good rationale for their use.
This chapter attempts to analyze the actual level of maturity and acceptance achieved by
intelligent systems and attempts to assess how, where and why they are (or can be) accepted
in the industry. Note that, although the focus is on
industrial applications
, this term generally
applies also to several other real-world applications such as agronomy, economics,
mathematics, weather forecasting, etc.
