The initial values of C and D are chosen at the center of
the normalized vector space. Then they are allowed to
change slightly in the vicinity of the original reference
vector. For each combination of vector C and D, the speed
error is noted and the combination that generates the least
error shall be declared the winner and takes over as the
reference vector pair for the next cycle of learning. The
new cycle of learning will be exactly the same as the
previous cycle. Learning shall stop when the speed error
has reached a tolerable value. The weights of the neural
network shall be taken from the final values of C and D.
111 THE FUZZY LOGIC MODEL
The fuzzy logic model proposed uses the speed error,
e(k), as a direct input and from this information, the rate
of change of error de(k), which is the derived input, is
computed. Fig. 3 shows the fuzzification process.