Goal of the research activity, carried out in the context of a
collaboration with Co.Ri.M.Me., is the analysis and the realization of
a hardware and software integrated system
exploiting the elaboration capabilities offered by fuzzy
logic.
The analysis of the software methodologies resulted in the
realization of an automatic synthesis environment for fuzzy
systems that includes algorithms for the interaction between the
information coming from natural language and from the lowest levels of
abstraction and algorithms that minimize the cost of the systems that
use this information.
A programmable analog fuzzy processor has been realized, which
processes up to 15 rules with three inputs and one
output. Starting from specifications of the desired input/output relationship
in terms of fuzzy rules or data samples, the
software calculates the programming values that must be
applied to the circuit in order to minimize the mean square error
between the specified and synthesized characteristic.
A digital bit-stream, serially loaded at system start-up from
an external EEPROM, programs the chip.
A compiler for analog fuzzy controllers has been also designed. The layout is
generated automatically starting from global system specifications and
the area and power requirements by any specific application are optimized.