Hello everyone,
I am wondering if anyone knows if there is any theoretical
analysis of the following problem:
Given a LTI continuous model, its system matrices(A,B)'s structure
are known. In saying so, I mean the dimension, the location of zeor
entries in the matrix. But not all the nonzero entries are known. Also
known are the input (u) and a state measurement say, x1.
By fitting the model to the u/x1 pair, one obtains A_hat and B_hat,
which are not necessarily equal to (A,B). Then the question is under
what condition, other states simulated with A_hat and B_hat will be
close to their true values given x1_hat close to x1.

Thanks!

xiao

Thanks!

xiao