Filtering and System Identification: A Least Squares Approach

9.1 Consider the system
where u(k) ? ? m and y(k) ? ? m with m>1. We are given the sequences of input-output data pairs
such that
with e i being the ith column of the n n identity matrix. Assume that N ? s, and show how the output data sequences
for i=1, 2, , m must be stored in the matrix
such that
9.2 Consider the subspace identification method for impulse input signals, described in Section 9.2.3.
Write a Matlab program to determine the system matrices A, B, and C up to a similarity transformation.
Test this program using 20 data points obtained from the following system:
Check the eigenvalues of the estimated A matrices, and compare the outputs from the models with the real output of the system.
9.3 Consider the minimal and asymptotically stable LTI system
with x(0)=0 and u(k) equal to a step sequence given by
Show that the data equation (9.7) on page 296 can be written as
where
j ? ? j denotes the vector with all entries equal to unity.
Show that
Show that
and use it to prove that
Use the results derived above to prove that
9.4 When the input to the state-space system
is periodic, the output will also be periodic. In this case there is no need to build a block Hankel...