Experiment design in Nonlinear Set Membership identification

Experiment design for nonlinear systems is considered within a set membership framework. A quantity tau
_I
, called radius of information, providing the worst-case identification error, is introduced. The radius of information is used to choose the most suitable experimental setting for identification. This approach allows to solve several experiment design problems. However, the computation of tau
_I
is difficult because it involves the evaluation of a function norm in a multi-dimensional space. Two algorithms are proposed: The first provides the exact value of tau
_I
but requires a computational complexity which is exponential in the dimension of the regressor space. The second provides an approximate value of tau
_I
and involves a polynomial (quadratic) complexity.