Information Technology Reference
The success of this applications of intelligent systems with respect to the previous system which was
based on heuristics depends on the following reasons: i) a the correct formalisation of the MOO
problem; ii) a suitable simulation system of the automatic warehouse (Colla & Nastasi 2010) that has
been realised in order to reproduce the monitoring system of the warehouse itself and can be fed with
the same data files that are used by the real system; iii) the possibility (as a consequence of point ii) to
easily test the different strategies in a realistic way without affecting the normal operations of the
warehouse; iv) the possibility (as a consequence of point ii) of performing an easy and user-friendly
comparison of the standard and simulated situation of the warehouse obtained through the previous
and improved strategies is possible, which can be of help for the technical personnel in order to
evaluate the advantages of a new strategy; v) the easiness of collecting training data, which are no
more than standard system data; vi) the modularity of the software for simulation and for the
implementation of the allocation strategy, which makes the substitution of the new code within the
control system of the warehouse straightforward.
According to authors' personal experience, it cannot be stated that intelligent systems are so
advantageous with respect to traditional techniques to be universally accepted for industrial
applications. Or better, advantages exist but they are often too limited when compared with
the additional risks, training costs, design time, and documentation/maintenance effort.
There are surely applications were they provide advantages, especially in tough problems,
but these are rather limited, therefore they do not justify a universal acceptance.
Unfortunately, in most industrial applications that we have encountered so far, very few
intelligent system offered such better performance with respect to other techniques to
really convince the sceptical user. Of course the comparison is made between the best
intelligent system and the correspondingly best non-intelligent technique.
5.1 A global advantage of intelligent system
There is perhaps a major advantage which makes intelligent systems attractive in a wider
range of applications. In practice, intelligent systems are:
generic approximation and modelling techniques which allow accurate system modelling/
forecasting/approximation/classification/etc. without any specific experience of the designer .
In practice anybody without any experience in a specific subject can afford solving a
problem which could otherwise (namely, with traditional techniques) be afforded only by
an expert (or a team of experts) in that field. It is likely that an expert, with appropriate
knowledge of the problem and of a bunch of more specific methods would achieve a better
result, but this would be far more expensive for an industry, both because of the higher cost
of the expert and for the longer development time. This is a rather interesting advantage,
even when intelligent systems are suboptimal, as it significantly reduces training costs of
5.2 How to help industry accepting intelligent systems
The authors personally believe that industry strongly needs to be helped to accept
intelligent systems and this should be a major role for universities and research institutions.